DETAILED ACTION
Remarks
Applicant presents a communication dated 26 May 2026 responsive to the 25 February 2026 non-final Office action (the “Previous Action”).
Claims 1, 3-5, 61 and 8-20 are amended. Claim 2 is cancelled.
Claims 1 and 3-20 remain pending. Claims 1, 12 and 16 are the independent claims.
Any unpersuasive arguments are addressed in the “Response to Arguments” section below.
Notice of Pre-AIA or AIA Status
The present application, filed on or after March 16, 2013, is being examined under the first inventor to file provisions of the AIA .
37 C.F.R. § 1.121
Applicant’s claim listing is not compliant with 37 C.F.R. § 1.121 because claim 6 is listed as “Original” when it has in fact, been amended. The claims are nonetheless examined in the interests of compact prosecution.
Examiner Notes
Examiner cites particular columns, paragraphs, figures and line numbers in the references as applied to the claims below for the convenience of the applicant. Although the specified citations are representative of the teachings in the art and are applied to the specific limitations within the individual claim, other passages and figures may apply as well. It is respectfully requested that, in preparing responses, the applicant fully consider the references in their entirety as potentially teaching all or part of the claimed invention, as well as the context of the passage as taught by the prior art or disclosed by the examiner.
In the event the determination of the status of the application as subject to AIA 35 U.S.C. 102 and 103 (or as subject to pre-AIA 35 U.S.C. 102 and 103) is incorrect, any correction of the statutory basis for the rejection will not be considered a new ground of rejection if the prior art relied upon, and the rationale supporting the rejection, would be the same under either status.
Response to Arguments
Applicant argues with respect to claim 1 that the cited references do not teach or reasonably suggest “one or more computer services employing the plurality of processors that define the AI workbench [and] further defines a machining-operations (MLOps) unit configured to provide real-time interactive testing of sample input values for the AI workflow, while the AI workflow is being actively assembled” because, according to Applicant, nowhere in the cited references is there any mention of users having the ability to receive real-time interactive feedback from an AI workbench, while actively deigning/developing an AI workflow. (Remarks, p. 6 par. 4).
Examiner respectfully disagrees for the reasons set forth in the rejections below. Applicant provides little more than a conclusion to the contrary.
Applicant’s arguments with respect to the remaining claims by virtue of their dependence from claim 1, similarity with claim 1 or dependence from a similar claim are unpersuasive for the same reasons.
Claim Interpretation
In view of Applicant’s amendments to the claims, no limitation of any claim is interpreted in accordance with 35 U.S.C. § 112(f).
Claim Rejections - 35 USC § 112
The Previous Action’s § 112 rejections are withdrawn in view of Applicant’s amendments and remarks unless reproduced herein below.
The following is a quotation of 35 U.S.C. 112(b):
(b) CONCLUSION.—The specification shall conclude with one or more claims particularly pointing out and distinctly claiming the subject matter which the inventor or a joint inventor regards as the invention.
The following is a quotation of 35 U.S.C. 112 (pre-AIA ), second paragraph:
The specification shall conclude with one or more claims particularly pointing out and distinctly claiming the subject matter which the applicant regards as his invention.
Claims 8-9 are rejected under 35 U.S.C. 112(b) or 35 U.S.C. 112 (pre-AIA ), second paragraph, as being indefinite for failing to particularly point out and distinctly claim the subject matter which the inventor or a joint inventor (or for applications subject to pre-AIA 35 U.S.C. 112, the applicant), regards as the invention.
As to claim 8, the claim refers to “the MLOps unit hosted on at least one server and employing at least one processor” at lines 2-3. There is insufficient antecedent basis for these limitations in the claims, as claim 1 only refers an “one or more computer servers employing the plurality of processors that define” an MLOps unit. For the purposes of examination, the “the MLOps unit hosted on at least one server and employing at least one processor” language will be interpreted as only referring back to the MLOps unit of claim 1 and not adding any additional limitations.
As to claim 9, the claim is dependent on claim 1 but does not cure the deficiencies of that claim. Accordingly, it is rejected for the same reasons.
Claim Rejections - 35 USC § 102
The following is a quotation of the appropriate paragraphs of 35 U.S.C. 102 that form the basis for the rejections under this section made in this Office action:
A person shall be entitled to a patent unless –
(a)(2) the claimed invention was described in a patent issued under section 151, or in an application for patent published or deemed published under section 122(b), in which the patent or application, as the case may be, names another inventor and was effectively filed before the effective filing date of the claimed invention.
Claims 1, 3-4, 6-7 and 11-15 are rejected under 35 U.S.C. 102(a)(2) as being anticipated by Du et al. (US 2025/0094137) (art of record – hereinafter Du).
As to claim 1, Du discloses: a computer-implemented method, executed on a computing device, comprising:
enabling a user, via an artificial intelligence (AI) workbench, to create an AI workflow; (e.g., Du, par. [0051]: the visual programming platform 30 [AI workbench, or comprising one] can include a node-graph editor; par. [0052]: user input 20 creates a node-graph that defines the computational pipeline; par. [0053]: users can add nodes, connect them with lines representing data flow; par. [0047]: nodes can include machine learning [a type of AI] algorithms) wherein the AI workbench is a full-stack platform defined by one or more computer servers employing a plurality of processors, configured to provide end-to-end functionality from a front-end user interface to a back-end combination of infrastructure and programming logic (e.g., Du, par. [0028]: the platform [workbench] can provide the graphical visualization of the pipeline front display in an interactive user interface [front-end user interface]; par. [0006]: the platform is implemented by one or more computing devices [infrastructure] and configured to perform operations; par. [0008]: instructions [programming logic], that when executed cause a computing system to perform operations [so because the platform includes a front-end and back-end combination of infrastructure and programming logic, it provides end-to-end functionality]; par. [0330]: processors 1103 can execute instructions 1106 and/or other instructions as described herein; par. [0037]: FIG. 12 depicts a cloud-based server system in accordance with an example embodiment. In FIG. 12, the functionality of a computing device [which implements the platform, see above] can be distributed amongst computing clusters).
converting the AI workflow into an executable computational representation; (e.g., Du, par. [0031]: users can arrange these components in the desired sequence to represent the flow of data through the pipeline [AI workflow]. The platform can then translate this graphical configuration into programming-language software code and vice-versa)
performing data collection, data processing, and data analysis for the AI workflow; (e.g., Du, par. [0047]: nodes can include machine learning algorithms. Example nodes can facilitate data collection, and/or data pre and post-processing; par. [0308]: machine algorithms to recognize patterns in the training data)
developing one or more visualization tools for the AI workflow; (e.g., Du, par. [0047]: example nodes can facilitate visualization of model results; par. [0050]: the node library 40 can provide an interface for users to create and contribute their own custom nodes)
developing a UI for the AI workflow; (e.g., Du, par. [0073]: the node graph-editor can generate a graphical visualization of the computational pipeline [workflow]. The node graph editor can provide the graphical visualization of the pipeline for display in the editor interface)
selecting one or more machine-learning (ML) models to be used in the AI workflow; (e.g., Du, par. [0249]: an example node-graph editor that enables users to select one or more nodes to generate a node-graph; par. [0048]: nodes that represent distinct machine learning models) and
developing a set of rules for interpreting results generated for each ML model of the one or more selected ML models (e.g., Du, par. [0258]: the menu of output formats may enable a user to select one or more output formats [each selected output representing a rule in the sense that the selection indicates that output should be in this format]) wherein the one or more computer servers employing the plurality of processors that define the AI workbench (see above) further defines a machine-learning operations (MLOps) unit configured (see above, instructions cause a computing system to perform operations. The instructions causing a computing system to performing the following are the “MLOps unit”) to provide real-time interactive testing of sample input values for the AI workflow, while the AI workflow is being actively assembled (e.g., Du, par. [0045]: the visual programming platform 30 can include testing and debugging tools. These can components can provide a unified environment for developers to test applications efficiently; Fig. 4D and associated text, par. [0287]: For example, a researcher may apply adjustments to a live image [sample input values] to test the robustness of two models side-by-side in real time. The options for adjustments 445 may be displayed when the image/video adjustment node 475 is displayed in the node graph editor [i.e., while the AI workflow is being actively assembled, see figure]; par. [0305]: for a given input image [sample input values], the blurriness may be changed in real-time, to test whether the models under test work in bright sunshine. As previously described with reference to FIG. 5, the pipeline may be tested on images from a video live stream. Prototyping platform 800 enables a user to drag and drop the camera [actively assemble the workflow, see figure 5] and view results from the ML models in real time).
As to claim 3, Du discloses the computer-implemented method of claim 1 (see rejection of claim 1 above), Du further discloses:
wherein the AI workbench includes a no-code user interface (UI) (e.g., Du, par. [0028]: the visual programming platform can provide the visualization of the pipeline for display in an interactive user interface that allows a user to edit the graphical visualization) hosted on at least one server and employing at least one processor configured (e.g., Du, par. [0006]: the platform is implemented by [hosted on] one or more computing devices and configured to perform operations; par. [0326]: computing device 1100 may include one or more processors; par. [0037]: FIG. 12 depicts a cloud-based server system in accordance with an example embodiment. In FIG. 12, the functionality of a computing device can be distributed amongst computing clusters) to allow users without coding skills (e.g., Du, par. [0002]: in traditional development, code is written in a text-based programming language; par. [0004]: in recent years, visual programming as emerged as an alternative to text-based programming; par. [0006] one example aspect of the present disclosure is directed to a visual programming platform) to produce customized AI tools, (e.g., Du, par. [0053]: users can add nodes, connect them with lines representing data flow; par. [0047]: nodes can include machine learning [a type of AI] algorithms) and to perform rapid prototyping (e.g., Du, par. [0045]: the visual programming platform can be referred to as a Rapid Application Prototyping System for Artificial Intelligence) without software development cycles (e.g., Du, par. [0265]: edits made in the node-graph may be reflected in the output without a need for code compilation, packaging and/or redeployment).
As to claim 4, Du discloses the computer-implemented method of claim 3 (see rejection of claim 3 above), Du further discloses:
wherein the no-code UI includes a workflow canvas (e.g., Du, par. [0052]: the node-graph editor 50 can be responsive for providing an interactive canvas in a user interface) hosted on at least one server and employing at least one processor (e.g., Du, par. [0006]: the platform is implemented by one or more computing devices and configured to perform operations; par. [0326]: computing device 1100 may include one or more processors; par. [0037]: FIG. 12 depicts a cloud-based server system in accordance with an example embodiment. In FIG. 12, the functionality of a computing device can be distributed amongst computing clusters) configured to allow users to select and reorganize one or more modular components when creating the AI workflow, (e.g., Du, par. [0052]: node-graph editor can receive user input 20 that creates or modifies a node-graph [AI workflow, see above] that defines the computational pipeline; par. [0053]: users can drag or drop nodes [modular components], connected them in a desired sequence; par. [0253]: the node-graph is edited “(e.g., edit each node, change a connecting edge between two nodes, and so forth)”) and a natural language programming (NLP) interface (see the citations for this claim below) hosted on at least one server and employing at least one processor (e.g., Du, par. [0006]: the platform is implemented by one or more computing devices and configured to perform operations; par. [0326]: computing device 1100 may include one or more processors; par. [0037]: FIG. 12 depicts a cloud-based server system in accordance with an example embodiment. In FIG. 12, the functionality of a computing device can be distributed amongst computing clusters) configured to receive and translate user-generated text-based, non-coded instructions defining a set of operational behaviors for each modular component of the one or more modular components selected for the AI workflow into machine-executable code-based instructions. (e.g., Du, par. [0065]: the platform 30 can initialize code 80. All the user needs to do is provide a natural language description of the pipeline “(e.g., in a text entry field of the platform’s user interface)”; par. [0073]: the node graph-editor can generate a graphical visualization of the computational pipeline defined by the set of code 80. For example., the node-graph editor can map each instruction “(e.g., nodes)” from the code 80 to graphical components. The node graph editor can provide the graphical visualization of the pipeline for display in the editor interface)
As to claim 6, Du discloses the computer-implemented method of claim 4 (see rejection of claim 1 above), Du further discloses:
wherein the NLP interface includes one or more large language models (LLMs) (e.g., Du, par. [0067]: platform 30 can process the natural language description with the machine learning coding system to generate a set of pseudocode. The coding system 90 can include one or more machine learned models which may be called generative sequence processing models. One example of a generative sequencing processing model is a so-called “large language model”) trained in a plurality of topics, including at least: machine learning concepts; data transformations, data serialization formats; and workflow schema templates (e.g., Du, Fig. 3 and associated text, par. [0241]: training a pseudocode drafting model; par. [0242]: the process can begin by obtaining a plurality of training tuples [machine learning concepts]. Training tuple 3002 can include a natural language description 3004 and a set of programming-language code 3006; par. [0240]: pseudocode 2012 can be processed by a compiler to generate a set of programming language code. The following is an example compiler output expressed in JSON [a data serialization format]; par. [0245]: the natural language description 3004 can be provided to the pseudocode drafting model 2010. The model 2010 can output a set of predicted pseudo code 3012 [i.e., transforming 3004 to 3012]; par. [0246]: A loss function 3014 [machine learning concept] can 3014 can evaluate whether the predicted pseudocode contains the same sequence as the ground truth pseudocode 3010 [the ground truth pseudocode being a workflow schema template at least because it is used as a pattern that the predicted pseudocode must match]).
As to claim 7, Du discloses the computer-implemented method of claim 4 (see rejection of claim 4 above), Du further discloses: wherein the workflow canvas includes:
a debugging window configured to provide oversight of every step of the information flow through the AI workflow; (e.g., Du, par. [0267] GUI 100 may be configured to support debugging in general. For example, users may tune parameters on any node [step of the information flow through the AI workflow]) and
a live-preview window configured to provide real-time rendering of the AI workflow while user changes are actively being made on the workflow canvas (e.g., Du, Fig. 4A and associated text, par. [0253]: enabling a user to update the graph by adding or removing a node or edge [parts of the workflow]. Such embodiments involve updating the output in real time. For example, output image 135 may be displayed [in the top section of the screen i.e., a panel, see figure] and real-time changes to output image 135 may be visitable to a user as the node-graph is edited).
As to claim 11, Du discloses the computer-implemented method of claim 1 (see rejection of claim 1 above), Du further discloses:
wherein the AI workflow includes an AI-powered assistant configured to recommend templates, provide contextual help, make suggestions, and guard against one or more misconceptions held by users when implementing machine learning processes (e.g., Du, par. [0280]: predicting, by a trained graph predictive model [AI-powered assistant], one or more of a next node or a next edge of the graph. Such embodiments involve recommending [suggesting] the one or more of the next node or the next edge to a user. For example, if a current node is an image, a next node recommendation may include one or more image processing models [such a recommendation being contextual help because it is based on a current node (context)]; par. [0271]: nodes may include miscellaneous items 225 “(e.g., a nose position extractor, HTML text, template, etc.)” [so recommended nodes can be templates because nodes can be templates]; par. [0279]: if a user attempts to connect two nodes that cannot be connected, process 300 may generate an error message [recommended nodes or edges would thus guard against misconceptions of the user about which nodes are able to be connected]).
As to claim 12, Du discloses a computer program product residing on a non-transitory computer readable medium having a plurality of instructions stored thereon which, when executed by a processor, cause the processor to perform operations (e.g., Du, par. [0006]: the platform is implemented by one or more computing devices configured to perform operations; par. [0325]: FIG. 11 is an example computing device 100 in accordance with embodiments; par. [0330]: processors 1103 can be configured to execute instructions 1106 in data storage 1104 and/or other instructions described herein; par. [0331]: data storage 1104 can includes one or more non-transitory computer-readable storage media) comprising:
creating, via an artificial intelligence (AI) workbench, an AI workflow; (e.g., Du, par. [0051]: the visual programming platform 30 [AI workbench, or comprising one] can include a node-graph editor; par. [0052]: user input 20 creates a node-graph that defines the computational pipeline; par. [0053]: uses can add nodes, connect them with lines representing data flow; par. [0047]: nodes can include machine learning [a type of AI] algorithms)
converting the AI workflow into an executable computational representation; (e.g., Du, par. [0031]: users can arrange these components in the desired sequence to represent the flow of data through the pipeline [AI workflow]. The platform can then translate this graphical configuration into programming-language software code and vice-versa)
performing data collection, data processing, and data analysis for the AI workflow; (e.g., Du, par. [0047]: nodes can include machine learning algorithms. Example nodes can facilitate data collection, and/or data pre and post-processing; par. [0308]: machine algorithms to recognize patterns in the training data)
enabling a user, via the AI workbench, to develop one or more visualization tools for the AI workflow; (e.g., Du, par. [0047]: example nodes can facilitate visualization of model results; par. [0050]: the node library 40 can provide an interface for users to create and contribute their own custom nodes)
enabling a user, via the AI workbench, to develop a UI for the AI workflow; (e.g., Du, par. [0065]: the platform 30 can initialize code 80. All the user needs to do is provide a natural language description of the pipeline. par. [0073]: the node graph-editor can generate a graphical visualization of the computational pipeline defined by the set of code 80. The node graph editor can provide the graphical visualization of the pipeline for display in the editor interface)
enabling a user, via the AI workbench, to select one or more machine-learning (ML) models to be used in the AI workflow; (e.g., Du, par. [0249]: an example node-graph editor that enables users to select one or more nodes to generate a node-graph; par. [0048]: nodes that represent distinct machine learning models) and
enabling a user, via the AI workbench, to develop a set of rules for interpreting results generated for each ML model of the one or more selected ML models (e.g., Du, par. [0258]: the menu of output formats may enable a user to select one or more output formats [each selected output representing a rule in the sense that the selection indicates that output should be in this format])
enabling a user, via the AI workbench, to provide real-time interactive testing of sample input values for the AI workflow, while the AI workflow is actively being assembled (e.g., Du, par. [0045]: the visual programming platform 30 can include testing and debugging tools. These can components can provide a unified environment for developers to test applications efficiently; Fig. 4D and associated text, par. [0287]: For example, a researcher may apply adjustments to a live image [sample input values] to test the robustness of two models side-by-side in real time. The options for adjustments 445 may be displayed when the image/video adjustment node 475 is displayed in the node graph editor [i.e., while the AI workflow is being actively assembled, see figure]; par. [0305]: for a given input image [sample input values], the blurriness may be changed in real-time, to test whether the models under test work in bright sunshine. As previously described with reference to FIG. 5, the pipeline may be tested on images from a video live stream. Prototyping platform 800 enables a user to drag and drop the camera [actively assemble the workflow, see figure 5] and view results from the ML models in real time).
As to claim 13, Du discloses the computer program product of claim 12 (see rejection of claim 1 above), Du further discloses:
where the AI workbench is a full-stack platform defined by one or more computer services employing a plurality of processors, configured to provide end to end functionality form a front-end user interface to a back-end combination of infrastructure and programming logic (e.g., Du, par. [0028]: the platform [workbench] can provide the graphical visualization of the pipeline front display in an interactive user interface [front-end user interface]; par. [0006]: the platform is implemented by one or more computing devices [infrastructure] and configured to perform operations; par. [0008]: instructions [programming logic], that when executed cause a computing system to perform operations [so because the platform includes a front-end and back-end combination of infrastructure and programming logic, it provides end-to-end functionality]; par. [0037]: FIG. 12 depicts a cloud-based server system in accordance with an example embodiment. In FIG. 12, the functionality of a computing device [a computing device implements the platform, see above] can be distributed amongst computing clusters).
As to claim 14, it is a computer program product claim whose limitations are substantially the same as those of claim 3. Accordingly, it is rejected for substantially the same reasons.
As to claim 15, it is a computer program product claim whose limitations are substantially the same as those of claim 4. Accordingly, it is rejected for substantially the same reasons.
Claim Rejections - 35 USC § 103
The following is a quotation of 35 U.S.C. 103 which forms the basis for all obviousness rejections set forth in this Office action:
A patent for a claimed invention may not be obtained, notwithstanding that the claimed invention is not identically disclosed as set forth in section 102, if the differences between the claimed invention and the prior art are such that the claimed invention as a whole would have been obvious before the effective filing date of the claimed invention to a person having ordinary skill in the art to which the claimed invention pertains. Patentability shall not be negated by the manner in which the invention was made.
Claim 5 is rejected under 35 U.S.C. 103 as being unpatentable over Du (US 2025/0094137) in further view of Pillai et al. (US 2010/0077260) (art of record – hereinafter Pillai).
As to claim 5, Du discloses the computer-implemented method of claim 4 (see rejection of claim 4 above), Du further discloses:
wherein the one or more modular components are selected from a set of different component types, (e.g., Du, par. [0249]: users select one or more nodes to generate a node-graph; par. [0047]: nodes can include nodes that perform various operations) wherein the set of different component types includes at least:
a data input component (see below) hosted on at least one server and employing at least one processor (e.g., Du, par. [0006]: the visual programming platform is implemented by [hosted on] one or more computing devices [since the platform is hosted by the devices, all of its components are]; par. [00326]: computing device 1100 may include one or more processors 1103; par. [0330]: one or more processors 1103 can be configured to execute instructions in storage 1104 and/or other instructions described herein; par. [0037]: FIG. 12 depicts a cloud-based server system in accordance with an example embodiment. In FIG. 12, the functionality of a computing device can be distributed amongst computing clusters) configured to receive input data; (e.g., Du, par. [0047], as an example, the nodes can include input nodes; par. [0251]: for example, first node, image 1 140, may include a portion where a user may upload an image; par. [0256]: “the input node (e.g., first node 140)” [note that the platform is implemented; par. [0326]: computing device 100 may include data storage 1104;)
a pre-processing component hosted on at least one server and employing at least one processor configured to clean, reformat, encrypt, or transform input data; (e.g., Du, par, [0047]: nodes can include data transformation [and all components of Du are hosted on a server employing one or more processors, see the mapping of the “data input component” of this claim above])
a model component hosted on at least one server and employing at least one processor configured to run one ML model from a set of different ML models; (e.g., Du, par. [0048]: some nodes can execute pre-trained machine learning models such as BERT or ResNet [and all components of Du are hosted on a server employing one or more processors, see the mapping of the “data input component” of this claim above])
a logic-control component hosted on at least one server and employing at least one processor configured to implement one or more logic commands, including filtering (e.g., Du, par. [0292]: low-pass filter JS nodes [and all components of Du are hosted on a server employing one or more processors, see the mapping of the “data input component” of this claim above])
a data output component hosted on at least one server and employing at least one processor configured to send output data to one output format from a set of different output formats; (e.g., Du, par. [0278]: output node 325 may provide a user with an ability to select from various options. As illustrated, the user may select an output format to be “JSON viewer” [and all components of Du are hosted on a server employing one or more processors, see the mapping of the “data input component” of this claim above])
a visualization component hosted on at least one server and employing at least one processor configured to display intermediate results or final results in one display format from a set of different display formats; (e.g., Du, par. [0271]: node may include output options 220 “(e.g., MobileNet results, pose detection visualizer, JSON viewer, etc.)” [and all components of Du are hosted on a server employing one or more processors, see the mapping of the “data input component” of this claim above]) and
an annotation component hosted on at least one server and employing at least one processor configured to allow users to manually enter or modify labels or results; (e.g., Du, Fig. 4D and associated text, par. [0287]: a researcher may apply adjustments 445 “(e.g., rotation, scaling, offset)” to a live image. The results 450 of the image/video adjustments 445 may be displayed when adjustment node 475 [annotation component] is generated; par. [0055]: node inspector 70. It can be implemented as a panel that displays properties of the currently selected node. Nodes inspector 70 can provide various controls such as text fields, sliders, checkboxes and drop-down lists [and see figure, the use can modify a title (label), translate X, translate Y, rotation, etc. (results). The node (component) having these properties thus allows the users to modify them. Note too that all components of Du are hosted on a server employing one or more processors, see the mapping of the “data input component” of this claim above])
a monitoring component hosted on at least one server and employing at least one processor configured to collect and display performance metrics in one display format from the set of different display formats (e.g., Du, par. [0271]: node may include output options 220 “(e.g., MobileNet results, pose detection visualizer, JSON viewer, etc.)”; Fig. 4B and associated text, par. [0273]: MobileNet 235 may identify the dog as a “brittany spaniel” with a confidence score of “39.2%”, as a “golden retriever” with a score of “17.9%” [in a display, see figure. Note too that all components of Du are hosted on a server employing one or more processors, see the mapping of the “data input component” of this claim above])
Du does not explicitly disclose a logic-control component hosted on at least one server and employing at least one processor configured to implement one or more logic commands, including at least if-else statements, branching, and looping.
However, in an analogous art, Pillai discloses:
a logic-control component hosted on at least one computing device and employing at least one processor configured to implement one or more logic commands, including at least if-else statements, branching, and looping; (e.g., Pillai, par. [0014]: to create a test procedure represented by connected graphical icons representing activities [components]; par. [0032]: elements which control test procedure flow include an “If Else Condition” activity and a “While Loop” activity. The “If Else Condition” creates two possible paths [branches] based on a conditional expression that evaluates to True or False; par. [0024]: test procedure generator and controller 23 will be described in the context of instructions, such as program modules, being executed by a computer [i.e., a processor of the computer]; par. [0042]: the test procedure editor module 53 for creating the test procedure can be located in memory 33 and accessible by processor 27)
It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to modify the server computing device hosting various components taught by Du such that the components hosted on the device include a logic-control component configured to implement one or more logic commands, including at least if-else statements, branching, and looping, as taught by Pillai, as Pillai would provide the advantage of a means of conditionally executing different flow paths, and repeatedly performing operations in the flow. (See Pillai, pars. [0032], [0034]).
Claims 8-9 are rejected under 35 U.S.C. 103 as being unpatentable over Du (US 2025/0094137) in further view of Edelstein et al. (US 2020/0050936) (art of record – hereinafter Edelstein) and Fedoruk et al. (US 2025/0378386) (art of record – hereinafter Fedoruk).
As to claim 8, Du discloses the computer-implemented method of claim 1 (see rejection of claim 1 above), Du further discloses: wherein the MLOps unit hosted on at least one server and employing at least one processor (see rejection of claim 1 above, the instructions that perform the recited functions are the MLOps unit) is further configured (see below) to:
deploy one or more ML models selected for the AI workflow; (e.g., Du, par. [0045]: platform 30 can include deployment tools. These components can offer an environment for developers to deploy applications; par. [0046]: nodes that users can incorporate into their applications [i.e., the workflows of Du are applications]) and
models selected for the AI workflow (see rejection of claim 1 above).
Du does not explicitly disclose to trigger model retraining for ML models selected for the AI workflow; monitor and report on performance metrics for the AI workflow; and perform versioning and generate audit trails for the AI workflow.
However, in an analogous art, Edelstein discloses to:
trigger model retraining for ML models (e.g., Edelstein, par. [0051]: retraining a deep learning model used by the software application).
It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to modify the MLOps unit and ML models selected for an AI workflow of Du such that the unit is configured to trigger retraining for the ML models, as taught by Edelstein, as Edelstein would provide the advantage of a means of improving the model. (See Edelstein, par. [0027]).
Further, in an analogous art, Fedoruk discloses: a machine-learning operations (MLOps) unit configured (e.g., Fedoruk, par. [0009]: users can use the platform to build AI agents; par. [0395]: an embodiment combining software and hardware may generally be referred to herein as a module. Aspects may take the form of a computer program product [so whatever code performs the following is the MLOps unit]) to:
monitor and report on performance metrics for the AI workflow; (e.g., Fedoruk, par. [0013]: the AI agent can be tested in a simulated execution; par. [0014]: the administrated can identify at least one execution metric to monitor as part of the simulated execution; par. [0316]: The results can include an output that displays onscreen. The test results are one type of execution metric “(also called ‘performance metric’)”. Other execution metrics can include cost and time to execute the agent [AI workflow]) and
perform versioning and generate audit trails for the AI workflow (e.g., Fedoruk, par. [0015]: the system can cause the AI agent [AI workflow] to be deployed. This can include indicating a version identifier of the tested agent is now the active version; par. [0062]: a next step “(e.g., an agent object)” of the AI agent’s workflow; par. [0011]: to generate manifest file based on selected agent objects that are connected on the UI; par. [0277]: manifest steps can include dates that track when the step was created and modified).
It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to modify the workbench of Du by incorporating a machine-learning operations (MLOps) unit configured to monitor and report on performance metrics for the AI workflow; and perform versioning and generate audit trails for the AI workflow, as taught by Fedoruk, as Fedoruk would provide a means to test the workflow and track versions of the workflow. (See Fedoruk, pars. [0015], [0107], [0014] and [0347]).
As to claim 9, Du/Fedoruk discloses the computer-implemented method of claim 8 (see rejection of claim 8 above) but Du does not explicitly disclose further including: deploying the one or more selected ML models, via the one or more computer servers employing the plurality of processors that define the MLOps unit, in a network environment; retraining the one or more ML models previously deployed in the AI workflow; and reintegrating the retrained ML models into the AI workflow.
However, in an analogous art, Fedoruk discloses further including:
deploying the one or more selected ML models, via the one or more computer servers employing the plurality of processors that define the MLOps unit, in a network environment; (e.g., Fedoruk, par. [0078]: an AI application service 113 executes on one or more processors 112 of the server device. The AI app service 113 can act as an AI platform that provides customers with a way to deploy AI agents; par. [0090]: AI agents can be deployed elsewhere, such as in a client application 145 [in network environment, see Fig. 1B of Fedoruk]; par. [0082]: objects within the agent, such as specific AI models 182) and
the one or more ML models previously deployed in the AI workflow (see immediately above).
It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to modify the workbench of Du by incorporating a machine-learning operations (MLOps) unit configured to: deploy selected models in a network environment, as taught by Fedoruk, as Fedoruk would provide a means of executing the models on a networked device. (See Fedoruk, Fig. 1B, par. [0090]).
Further in an analogous art, Edelstein discloses:
retraining the one or more ML models (e.g., Edelstein, par. [0051]: retraining a deep learning model used by the software application) and
reintegrating the retrained ML models into the AI workflow (e.g., Edelstein, par. [047]: a previously trained learning model is retrained to create an improved learning model. The improved deep learning model by be distributed back to the client system having the software application installed; abstract: models may be accessible to the software applications for providing inference data to the applications, which may then use as desired [a workflow, in the sense that a sequence of steps is performed, generating the inference and using it]).
It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to modify the MLOps unit and ML models selected for an AI workflow and deployed of Du/Fedoruk to include retraining the ML modes and reintegrating the retrained models into the AI workflow, as taught by Edelstein, as Edelstein would provide the advantage of a means of improving the workflow that uses the model. (See Edelstein, par. [0027]).
Claim 10 is rejected under 35 U.S.C. 103 as being unpatentable over Du (US 2025/0094137) in further view of Norwood et al. (US 2024/0370235) (art of record – hereinafter Norwood).
As to claim 10, Du discloses the computer-implemented method of claim 1 (see rejection of claim 1 above), Du further discloses the AI workbench (see rejection of claim1 aboe) but does not explicitly disclose wherein the AI workbench is configured to be scalably deployed in a plurality of network environments, including at least: cloud networks, on-premises networks, and edge-computing networks.
However, in an analogous art, Norwood discloses:
wherein the application is configured to be scalably deployed in a plurality of network environments, including at least: cloud networks, on-premises networks, and edge-computing networks (e.g., Norwood, par. [0007]: generating E2C application across public clouds, on-premise data centers and edge computing networks; par. [0003]: cloud computing environments provide an orchestration platform that make it easy to scale the application).
It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to modify the AI workbench of Du such that it is configured to be scalably deployed in a plurality of network environments, including at least: cloud networks, on-premises networks, and edge-computing networks, as taught by Norwood, as Norwood would provide the advantage of a means of splitting the computational logic across data centers and reducing bandwidth of network collected by the equipment. (See Norwood, par. [0004]).
Claim 16 is rejected under 35 U.S.C. 103 as being unpatentable over Fedoruk (US 2025/0378386) in view of Edelstein (US 2020/0050936).
As to claim 16, Fedoruk discloses an artificial intelligence (AI) workbench system hosted on one or more servers connected to a scalable network (e.g., Fedoruk, Fig. 1 and associated text, par. [0077]: a computer based-system 100 configured to implement one or more aspects of at least one embodiment. The system includes a server computer device 110. The server devices and computing device 140 are in communication over a network 130, such as the Internet [a scalable network]), comprising:
a no-code user interface (UI) hosted on at least one server and employing at least one processor (e.g., Fedoruk, par. [0078]: an artificial intelligence (“AI”) application (“app”) service (also called “AI platform”) executes on one or more processors of the server device; par. [0055]: the platform can present a user interface) configured to allow users without coding skills to produce customized artificial intelligence (AI) tools, (e.g., Fedoruk, par. [0205]: an AI agent can be designed using an exemplary UI. Using the menu section 520, a user can drag-and-drop objects from the meus and add relationships between such objects to design an AI agent [i.e., the agent is designed in this manner instead of writing code]; par. [0319]: the layman version avoids presenting information at a code level and instead provides concepts that get translated to code and parameters based on selections the user makes) and to perform rapid prototyping (e.g., Fedoruk, par. [0205]: objects in the AI agent 502. Menus for a user to select model objects for including in the agent 502 [the agent being a prototype until it is in production]; par. [0351]: categorization helps users quickly locate models suited to their specific needs) without software development cycles; (e.g., Fedoruk, par. [0004]: written program code may need to be repeatedly re-written. Currently there are few, if any, good ways to avoid re-writing program code to take advantage of newer models [so the Fedoruk avoids this because Fedoruk does not require writing any code]; par. [0310]: the process can simulate the agent execution to identify any issues before actual deployment [deployment being part of a development cycle, so a prototype can be created and run without a full cycle]) and
a machine-learning operations (MLOps) unit (see below, whatever portions of software perform the recited functions are the MLOp unit) hosted on at least one server and employing at least one processor (par. [0078]: an artificial intelligence (“AI”) application (“app”) service (also called “AI platform”) executes on one or more processors of the server device) configured to:
deploy ML models selected for an AI workflow, (e.g., Fedoruk, par. [0055]: the platform can present a user interface (UI) with agent objects and tools for constructing an AI agent, which can display as a flow diagram on the UI; par. [0205]: menus for a user to select model objects for including in the agent 502; par. [0078]: an AI platform that provides customers with a way to create, deploy and manage AI agents 116)
ML models selected for the AI workflow, (e.g., Fedoruk, par. [0107]: feedback from this process may be used to retrain models [the models are selected as noted above])
monitor and report on performance metrics for the AI workflow (e.g., Fedoruk, par. [0013]: the AI agent can be tested in a simulated execution; par. [0014]: the administrated can identify at least one execution metric to monitor as part of the simulated execution; par. [0316]: The results can include an output that displays onscreen. The test results are one type of execution metric “(also called ‘performance metric’)”. Other execution metrics can include cost and time to execute the agent [AI workflow]) and
perform versioning and generate audit trails for the AI workflow (e.g., Fedoruk, par. [0015]: the system can cause the AI agent [AI workflow] to be deployed. This can include indicating a version identifier of the tested agent is now the active version; par. [0062]: a next step “(e.g., an agent object)” of the AI agent’s workflow; par. [0011]: to generate manifest file based on selected agent objects that are connected on the UI; par. [0277]: manifest steps can include dates that track when the step was created and modified)
provide real-time interactive testing of sample input values for the workflow, while the AI workflow is being actively assembled (e.g., Fedoruk, par. [0055]: the platform can present a user interface (UI) with agent objects and tools for constructing an AI agent, which can display as a flow diagram on the UI [i.e., the agent = workflow]; Fig. 12 and associated text, par. [0236]: at step 12-2, the AP platform application 115 receives one or more inputs for testing an AI agent; par. [0237]: at step 1204, the AI platform application processes the input(s) received via the AI agent that is being tested; par. [0239]: at step 120 the AI platform application 115 receives a user definition of an updated agent [i.e., the agent is being actively assembled]; par. 0240]” after the AI Platform updates the AI agent, the method 1200 returns to step 1202, where the AI platform can receive more inputs; par. [0223]: AI agents can run in real time. For example, one-to-ten test cases can be performed based on live input).
Fedoruk does not explicitly disclose to trigger model retraining for ML models.
However, in an analogous art, Edelstein discloses
to trigger model retraining for ML models (e.g., Edelstein, par. [0051]: retraining a deep learning model used by the software application).
It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to modify the MLOps unit and ML models selected for an AI workflow of Fedoruk such that the unit is configured to trigger retraining for the ML models, as taught by Edelstein, as Edelstein would provide the advantage of a means of improving the model. (See Edelstein, par. [0027]).
Claims 17 and 19-20 is rejected under 35 U.S.C. 103 as being unpatentable over Fedoruk (US 2025/0378386) in view of Edelstein (US 2020/0050936) in further view of Du (US 2025/0094137)
As to claim 17, it is a system claim having limitations substantially the same as those of claim 4. They are not explicitly taught by Fedoruk but are taught by Du for the reasons set forth above with respect to that claim.
It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to modify Fedoruk by incorporating a workflow canvas hosted on at least one server and employing at least one processor configured to allow users to select and reorganize one or more modular components when creating the AI workflow; and a natural language programming (NLP) interface hosted on at least one server and employing at least one processor configured to receive and translate user-generated text-based, non-coded instructions defining a set of operational behaviors for each modular component of the one or more modular components selected for the AI workflow into machine-executable code-based instructions, as taught by Du, as Du would provide the advantage of a means to support users to intuitively and rapidly generate a workflow, which can then be edited by a user. (See Du, pars. [0033], [0030]).
As to claim 19, it is a system claim having limitations substantially the same as those of claim 6. They are not explicitly taught by Fedoruk but are taught by Du for the reasons set forth above with respect to that claim.
It would have been obvious to modify the workbench of Fedoruk to include an NLP interface hosted one on at least one server and employing at least one processor that includes one or more large language models (LLMs) trained in a plurality of topics, including at least: machine learning concepts; model application programming interfaces (APIs); data transformations; data serialization formats; and workflow schema templates, as taught by Du, as Du would provide the advantage of a means to support users to intuitively and rapidly generate a workflow. (See Du, par. [0033]).
As to claim 20, Fedoruk/Edelstein discloses the AI workbench system of claim 16 (see rejection of claim 16 above but does not explicitly disclose: further including an AI-powered assistant hosted on at least one server and employing at least one processor configured to recommend templates, provide contextual help, make suggestions, and guard against one or more misconceptions held by users when implementing machine learning processes.
However, in an analogous art, Du discloses:
further including an AI-powered assistant (see below) hosted on at least one server and employing at least one processor (e.g., Du, par. [0005]: another example aspect of the present disclosure is directed to media that store instructions, that, when executed, cause a computing system to perform operations; par. [0006]: the platform is implemented by [hosted on] one or more computing devices; par. [0330]: processors 1103 can execute instructions 1106 and/or other instructions as described herein; par. [0037]: FIG. 12 depicts a cloud-based server system in accordance with an example embodiment. In FIG. 12, the functionality of a computing device can be distributed amongst computing clusters) configured to recommend templates, provide contextual help, make suggestions, and guard against one or more misconceptions held by users when implementing machine learning processes (e.g., Du, par. [0280]: predicting, by a trained graph predictive model [AI-powered assistant], one or more of a next node or a next edge of the graph. Such embodiments involve recommending [suggesting] the one or more of the next node or the next edge to a user. For example, if a current node is an image, a next node recommendation may include one or more image processing models [such a recommendation being contextual help because it is based on a current node (context)]; par. [0271]: nodes may include miscellaneous items 225 “(e.g., a nose position extractor, HTML text, template, etc.)” [so recommended nodes can be templates because nodes can be templates]; par. [0279]: if a user attempts to connect two nodes that cannot be connected, process 300 may generate an error message [recommended nodes or edges would thus guard against misconceptions of the user about which nodes are able to be connected]).
It would have been obvious to modify the workbench of Fedoruk to include an AI-powered assistant hosted on at least one server and employing at least one processor configured to recommend templates, provide contextual help, make suggestions, and guard against common misconceptions held by users when implementing machine learning processes, as taught by Du, as Du would provide the advantage of a means for users to more easily generate a workflow. (See Du, par. [0030]).
Claim 18 rejected under 35 U.S.C. 103 as being unpatentable over Fedoruk (US 2025/0378386) in view of Edelstein (US 2020/0050936) in view of Du (US 2025/0094137) in further view of Pillai (US 2010/0077260).
As to claim 18, it is a system claim whose limitations are substantially the same as those of claim 5. They are not explicitly taught by Fedoruk but they are taught by Du and Pillai for the reasons set forth above with respect to that claim.
It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to modify the components of Fedoruk to include modular components are selected from a set of different component types all hosted on at least one server and employ at least one processor configured to perform their functions, wherein the set of component types includes at least: a data input component configured to receive input data; a pre-processing component configured to clean, reformat, encrypt, or transform input data; a model component configured to run one ML model from a set of different ML models; a logic-control component configured to implement one or more logic commands, including filtering; a data output component configured to send output data to one output format from a set of different output formats; a visualization component configured to display intermediate results or final results in one display format from a set of different display formats; an annotation component configured to allow users to manually enter or modify labels or results; and a monitoring component configured to collect and display performance metrics in one display format from the set of different display formats, as taught by Du, as Du would provide the advantage of a means for users to easily design workflows. (See Du par. [0030]).
It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to modify the components of Fedoruk/Edelstein/Du to include a logic-control component configured to implement one or more logic commands, including at least if-else statements, branching, and looping, as taught by Pillai, as Pillai would provide the advantage of a means of conditionally executing different flow paths, and repeatedly performing operations in the flow. (See Pillai, pars. [0032], [0034]).
Conclusion
Applicant's amendment necessitated the new ground(s) of rejection presented in this Office action. Accordingly, THIS ACTION IS MADE FINAL. See MPEP § 706.07(a). Applicant is reminded of the extension of time policy as set forth in 37 CFR 1.136(a).
A shortened statutory period for reply to this final action is set to expire THREE MONTHS from the mailing date of this action. In the event a first reply is filed within TWO MONTHS of the mailing date of this final action and the advisory action is not mailed until after the end of the THREE-MONTH shortened statutory period, then the shortened statutory period will expire on the date the advisory action is mailed, and any nonprovisional extension fee (37 CFR 1.17(a)) pursuant to 37 CFR 1.136(a) will be calculated from the mailing date of the advisory action. In no event, however, will the statutory period for reply expire later than SIX MONTHS from the mailing date of this final action.
Any inquiry concerning this communication or earlier communications from the examiner should be directed to TODD AGUILERA whose telephone number is (571)270-5186. The examiner can normally be reached M-F 11AM - 7:30PM EST.
Examiner interviews are available via telephone, in-person, and video conferencing using a USPTO supplied web-based collaboration tool. To schedule an interview, applicant is encouraged to use the USPTO Automated Interview Request (AIR) at http://www.uspto.gov/interviewpractice.
If attempts to reach the examiner by telephone are unsuccessful, the examiner’s supervisor, Hyung S Sough can be reached at (571)272-6799. The fax phone number for the organization where this application or proceeding is assigned is 571-273-8300.
Information regarding the status of published or unpublished applications may be obtained from Patent Center. Unpublished application information in Patent Center is available to registered users. To file and manage patent submissions in Patent Center, visit: https://patentcenter.uspto.gov. Visit https://www.uspto.gov/patents/apply/patent-center for more information about Patent Center and https://www.uspto.gov/patents/docx for information about filing in DOCX format. For additional questions, contact the Electronic Business Center (EBC) at 866-217-9197 (toll-free). If you would like assistance from a USPTO Customer Service Representative, call 800-786-9199 (IN USA OR CANADA) or 571-272-1000.
/TODD AGUILERA/Primary Examiner, Art Unit 2192
1 See the “37 C.F.R. § 1.121” Section below
Prosecution Insights