DETAILED ACTION
This action is responsive to the Applicant’s response filed 6/08/26.
Following the pre-appeal conference decision per 6/18/26, prosecution is herein re-opened. No claims have been amended. Claims 1-3, 5-14, 16-22 are pending prosecution by the following office action.
Claim Rejections - 35 USC § 101
35 U.S.C. 101 reads as follows:
Whoever invents or discovers any new and useful process, machine, manufacture, or composition of matter, or any new and useful improvement thereof, may obtain a patent therefor, subject to the conditions and requirements of this title.
Claim 12 is/are rejected under 35 U.S.C. 101 because the claimed invention is directed to a judicial exception (i.e., a law of nature, a natural phenomenon, or an abstract idea) without significantly more. Claim(s) 12 is/are directed to Abstract Idea. The claim(s) does/do not include additional elements that are sufficient to amount to significantly more than the judicial exception because the 2-step analysis as follows.
Step 1: claim 12 is directed to a method/process category.
Step 2A
Prong one:
Per MPEP 2106.04(a), in claim 12, the steps of importing dataset, performing architecture search, selecting/training a model, generating a pipeline, and exporting fall under the Certain Mental Processes abstract idea grouping (concepts performable in the human mind or using pen and paper) and/or Mathematical Concepts / Organizing Human Activity.
Specifically, the limitations recited as
"importing a training dataset associated with a user"
"performing a model architecture search that selects and trains a machine learning model for the training dataset"
"generating a deployment machine learning pipeline..."
"exporting the machine learning model and the deployment machine learning pipeline” describe high-level steps of collecting data, analyzing/evaluating models, selecting an appropriate model architecture, and organizing the resulting model/pipeline for deployment. The high-level expression of concepts like selecting, training, and organizing mathematical models based on incoming training data is depictive of a process that can be performed conceptually in the human mind or on paper using logical reasoning, statistical analysis, and basic decision-making rules. MPEP 2106.04 (a). The inclusion of automated machine learning observed at this high level of generality represents use of generic computer tools, abstract administrative and mathematical concept of data organization and model selection. MPEP 2106.04(d)
Claim 12 is directed to a mental process/Mathematical concept type of Judicial Exception.
Prong two:
The elements recited as "one or more computing devices" for executing generic functions of importing, executing, generating and exporting do not modify (or cannot be seen as modifying) or improve the operational functioning of the computer hardware itself; rather, the computer is merely used as a generic tool to execute the abstract process faster or more efficiently. MPEP 2106.05(d)
The importing/exporting step is also construed as pre-activity, post-activity step deemed insignificant toward improving the Abstract Idea. MPEP 2106.05(g)
The mention of automated machine learning (for selection, training) observed from a high level of generality (as claimed) represents an abstract administrative and mathematical concept of data organization and model selection, absent detailed depiction on the selecting, training associated with this mathematical concept (i.e. machine learning) means. Linking a generically expressed mathematical means with use of a Judicial Exception cannot integrate the Abstract Idea into a practical application as no meaningful transformation is perceived. MPEP 2106.05(c ) ( e) (f)
Moreover, the claim fails to provide specific implementation details or hardware architecture modifications that improve the underlying system. Under MPEP § 2106.04(d), simply invoking a generic computer to execute an abstract data manipulation process does not integrate the exception into a practical application. MPEP 2106.05(f)
Step 2B
The additional elements of “one or more computing devices” “pipeline” “machine learning” are tools or concepts otherwise perceived as generic to a computing environment used for performing the abstract Idea (importing, searching generating, exporting). The mere aspect of performing well-understood, routine, and conventional computer processing steps, in all, does not provide significantly more (MPEP § 2106.05(d)) to the Abstract Idea of step 2A.
As an ordered combination, the steps simply dictate a standard, logical workflow for data processing (import data [Wingdings font/0xE0] evaluate/train model [Wingdings font/0xE0] generate pipeline [Wingdings font/0xE0] export). This sequence adds no unconventional techniques or technical transformations beyond the standard lifecycle of data processing.
The claim merely recites what result is achieved (producing and deploying a model automatically) rather than a specific technical how (e.g., specific hardware cache optimization, custom algorithm logic, or a tangible system-level improvement); and merely performing complex mathematical or statistical operations at scale on a computer does not remove it from the realm of abstract mathematical concepts.
Accordingly, claim 12, as a whole, does not amount to significantly more than the abstract idea itself and is rejected under 35 U.S.C. 101. MPEP 2106.05
Claims 1 and 20 is/are rejected under 35 U.S.C. 101 because the claimed invention is directed to a judicial exception (i.e., a law of nature, a natural phenomenon, or an abstract idea) without significantly more. Claim(s) 1 and 20 is/are directed to Abstract Idea. The claim(s) does/do not include additional elements that are sufficient to amount to significantly more than the judicial exception because the 2-step analysis as following
A. Eligibility of claim 1
Step 1: claim 1 is directed to system category.
Step 2A
Prong one:
The elements recited as
"importing a training dataset associated with a user"
"performing a model architecture search that selects and trains a machine learning model for the training dataset"
"generating a deployment machine learning pipeline..."
"exporting the machine learning model and the deployment machine learning pipeline..."
describe high-level steps of collecting data, analyzing/evaluating models, selecting an appropriate model architecture, and organizing the resulting model/pipeline (e.g. generating a model) for deployment, and outputting the result for deployment. The high-level expression of concepts like selecting, training, and organizing mathematical models based on incoming training data is depictive of a process that can be performed conceptually in the human mind or on paper using logical reasoning, statistical analysis, and basic decision-making rules. MPEP 2106.04(a)
The inclusion of automated machine learning observed at this high level of generality represents use of generic computer tools, abstract administrative and mathematical concepts of data organization and model selection. MPEP 2106.04(d)
Prong two:
The elements recited as "one or more computing devices" for executing generic functions of importing, executing, generating and exporting do not modify (or cannot be seen as modifying) in terms of improving the operational functioning of the computer hardware itself; rather, the computer device here is merely used as a generic tool to execute the abstract process faster or more efficiently- MPEP 2106.05(d)
The importing/exporting step is mostly construed as a pre-activity, post-activity step deemed insignificant toward improving the Abstract Idea. MPEP 2106.05(g)
The mention of automated machine learning (for selection, training) observed from a high level of generality (as claimed) represents an abstract administrative and mathematical concept of data organization and model selection, absent detailed depiction about the selecting, training associated with this mathematical concept (i.e. machine learning) means. Linking a generically expressed mathematical means with use of a Judicial Exception cannot integrate the Abstract Idea into a practical application. MPEP 2106.05(c) (e) (f)
Step 2B
The additional elements of “one or more computing devices” “pipeline” “machine learning” are tools or concepts otherwise perceived as generic to a computing environment used for performing the abstract Idea (importing, searching generating, exporting). The mere aspect of performing well-understood, routine, and conventional computer processing steps, in all, does not provide significantly more (MPEP § 2106.05(d)) to the Abstract Idea of step 2A.
As an ordered combination, the steps simply dictate a standard, logical workflow for data processing (import data [Wingdings font/0xE0] evaluate/train model [Wingdings font/0xE0] generate pipeline [Wingdings font/0xE0] export). This sequence adds no unconventional techniques or technical transformations beyond the standard lifecycle of data processing.
The claim merely recites what result is achieved (producing and deploying a model automatically) rather than a specific technical how (e.g., specific hardware cache optimization, custom algorithm logic, or a tangible system-level improvement); and merely performing complex mathematical or statistical operations at scale on a computer does not remove it from the realm of abstract mathematical concepts.
Accordingly, claim 1, considered as a whole, does not amount to significantly more than the abstract idea itself and is rejected under 35 U.S.C. 101. MPEP 2106.05
B. Eligibility of claim 20
Step 1: claim 20 is directed to medium/product category
Step 2A
Prong one:
This claim recites steps of "performing a model architecture search that selects and trains a machine learning model for the training dataset", "generating a deployment machine learning pipeline..." are construed as high-level expression of concepts like selecting, training, and organizing mathematical models based on incoming training data is depictive of a process that can be performed conceptually in the human mind or on paper using logical reasoning, statistical analysis, and basic decision-making rules. MPEP 2106.04(a)
The inclusion of automated machine learning observed at this high level of generality represents use of generic computer tools, abstract administrative and mathematical concept of data organization and model selection. MPEP 2106.04(d)
Prong two:
The elements recited as "one or more computing devices" for executing generic functions of importing, executing, generating and exporting do not modify (or cannot be seen as modifying) or improve the operational functioning of the computer hardware itself; rather, the computer is merely used as a generic tool to execute the abstract process faster or more efficiently- MPEP 2106.05(d)
The importing/exporting step is mostly construed as a pre-activity, post-activity step deemed insignificant toward improving the Abstract Idea. MPEP 2106.05(g)
The mention of automated machine learning (for selection, training) as expressed in a high level of generality represents an abstract administrative and mathematical concept of data organization and model selection, absent detailed depiction about the selecting, training associated with this mathematical concept (i.e. machine learning) means. Linking a generically expressed mathematical means with use of a Judicial Exception cannot integrate the Abstract Idea into a practical application. MPEP 2106.05(c) (e) (f)
Step 2B
The additional elements of “one or more computing devices” “pipeline” “machine learning” are tools or concepts otherwise perceived as generic to a computing environment used for performing the abstract Idea (importing, searching generating, exporting). The mere aspect of performing well-understood, routine, and conventional computer processing steps, in all, does not provide significantly more (MPEP § 2106.05(d)) to the Abstract Idea of step 2A.
As an ordered combination, the steps simply dictate a standard, logical workflow for data processing (import data [Wingdings font/0xE0] evaluate/train model [Wingdings font/0xE0] generate pipeline [Wingdings font/0xE0] export). This sequence adds no unconventional techniques or technical transformations beyond the standard lifecycle of data processing.
The claim merely recites what result is achieved (producing and deploying a model automatically) rather than a specific technical how (e.g., specific hardware cache optimization, custom algorithm logic, or a tangible system-level improvement); and merely performing complex mathematical or statistical operations at scale on a computer does not remove it from the realm of abstract mathematical concepts.
Accordingly, claim 20 considered as a whole, does not amount to significantly more than the abstract idea itself and is rejected under 35 U.S.C. 101. MPEP 2106.05
Step 2B analysis of dependent claims.
Claims 2 and 13 recite use of structured labels as training data and selection thereof for a prediction by a ML. Providing structured information into a machine learning tool does not technically convert the abstract Idea into a non-conventional and novel improvement to this computer field in which the abstract idea operates.
Claims 3 and 14 recites receiving user statement of a problem and performing inferring based thereon to provide parameters for optimizing a ML search. The claims merely recites a result without pointing out how the optimization or machine inference is being carried out. Mere recitation of a intended use or desired result cannot be seen as a concrete and non-conventional transformation bringing about a significant technical improvement to this computer field in which the abstract idea operates
Claims 5 and 16 recite detecting one or more semantic feature and constraining a search based thereon to obtain candidate models capable to process the semantic type. Use of a semantic element as criterion to search for a model amounts to standard use of criterion to support a search as integral part of a mental process using support of mathematical concepts (machine learning)
Claims 6 and 17 recite use of statistics as basis for setting prediction by a training; an use of a metric or statistic as input/setting to train a model amounts to standard use of criterion to support a mathematical concepts integrated with a mental processes (abstract Idea) using support of machine learning or numerical techniques, not adding significantly more to the Abstract Idea.
Claims 7 and 18 recite predicting setting of a model search based on statistical description and processing thereof to support the machine learning prediction. The well-understood setting of a mathematical concepts tool fails to upconvert the abstract idea into a concrete and novel transformation required of step 2B.
Claim 8 recites identifying based on statistics descriptive of a dataset, searches and defining setting thereof based on previous settings or searches. The well-understood acts of acquiring information to organize them in association with a Abstract Idea fail to upconvert the abstract idea into a concrete and novel transformation required of step 2B
Claim 9 recites metadata descriptive of performance of a search, and parameters tuning of based thereon in support for performance of the search. Use of descriptive metric or metadata as input/setting into a ML-based search amounts to standard use of criterion to support a mathematical concept integrated with a mental process (abstract Idea) having support of machine learning or numerical techniques, not adding significantly more to the Abstract Idea.
Claim 10 recites identifying correlations between labels of a training set and this falls into the standard organization of information associated with data processing typical to a Abstract Idea, hence fail to add significantly more to the Judicial exception
Claim 11 and 19 recite fixed aspect of training and a retuning aspect thereof in association with identifying a new ML model to correspond to a dataset; but iterative effect of a search cannot depicts how the abstract idea of searching is improved or solves a technical aspect of the computer field in which the ML-based search operates.
Claim 21 and 22 recite a request received from a user to cause a ML pipeline to execute. No part of the request depicts how the operations associated with the Abstract Idea upconvert to the Abstract idea more significantly than itself, and evaluating performance of the ML, validating if its output satisfies the request. No part of the recited machine learning based operations/pipelining is particularly depicted to bolster technical improvement in using a ML to support pipeline, and no part in evaluating and validating steps describes how this is particularly being carried out to technically improve the computer field in which the ML-based abstract Idea operates; thus, claims 21-22 fail to add significantly more to the Judicial Exception identified per step 2A.
In all, claims 1-3, 5-14, 16-22 are deemed non-eligible under the 35 USC § 101 statute.
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.
Claims 1-3, 5-14, 16-22 is/are rejected under § 35 U.S.C. 103 as being unpatentable over Mueller et al, USPubN: 2021/0326717 (herein Mueller) in view of Siracusa et al, TW 2021135577 (translation), 04-01-2021, 29 pgs.(herein Siracusa).
As per claim 1, Mueller discloses a computing system for automatic production of machine learning models (see Abstract) and corresponding deployment pipelines (see Abstract), the computing system comprising: one or more processors; and one or more non-transitory computer-readable media that collectively store instructions that, when executed by the one or more processors, cause the computing system to perform operations, the operations comprising:
importing (if a new training set is provided …orchestrator can initiate a ML pipeline … using an AutoML library 121, Import library, load training data, file_path = PATH-TO-TRAINING-DATASET – para 0050; import TabularPrediction as task predictor … file_path = PATH-TO-TRAINING-DATASET – para 0051; pipeline exploration …presented to users … user interface 800 … users can “run” via selecting a user interface element … exploration user interface includes a second code section 810 that defines ways to process a training dataset – para 0078-79; exploration user interface - Fig. 7-8 – Note1: exploration interface enabling users to specify library code to configure a predictor task, a training set path and editing a ML pipeline reads on importing a training set associated with a interacting user ) a training dataset associated with a user (para 0008, 0038; training dataset … user may provide and select it via interacting with the GUI – para 0039);
executing (see below) an origination machine learning pipeline to:
perform a model architecture search (pipeline recommender system – para 0059-0060) that selects (identify a promising and diverse set of feature processing pipelines to apply to the customer dataset along with code for … feature processing model… also recommend hardware (compute instance type and count) identifiers of which hyperparameters and parallelization should use … when exploring the pipeline – para 0059) and trains a machine learning model (para 0043-0045) for the training dataset; and
generate a deployment machine learning pipeline (ML orchestrator 115 … deploying a model/pipeline to … a hosting system – para 0053; deploy a model/pipeline – para 0055) for deployment of the machine learning model; and
export calls and inference code commands (para 0073-0074 - Note2: deployment module of the ML orchestrator 115 sending API calls or pipeline commands to a model hosting system or batch model system configured for hosting deployment of the model/pipeline reads on exporting ML calls/commands and pipeline inference code of the model/pipeline to a model hosting system or a batch inference system for deployment of the ML model/pipeline) of the machine learning model and the machine learning pipeline for deployment of the machine learning model (a hosting system – para 0053; ML orchestrator 115 … may send … API calls to a model hosting system 140 described … to “host” the pipeline; to a batch inference system to “host” the pipeline – para 0074; ML hosting service that is associated with the deployed model/pipeline – para 0089) with the deployment machine learning pipeline.
Mueller does not explicitly disclose export of ML calls/commands or inference code of the pipeline as
export the machine learning model and the deployment ML pipeline for deployment of the MO model and the deployment ML pipeline.
Siracusa discloses training module to develop batch of training data and scheduling resources under use by developers via machine learning tool by which the developer can provide input to the training model (pg. 10-11) where the ML tool can import test data in structured table format, Javascript or templates using a import module API to retrieve (from database – see pg. 12) training data for configuration by the developers (pg. 6-7), where training a model includes, resource scheduling, test data for the learning (pg. 12) and evaluation of the model, such that a trained model can be exported for use by other applications (bottom pg 3), using a drag icon of a import/export module in a IDE to export the trained model destined for export(pg. 13). Hence, use of a export module in a machine learning tool IDE to export a trained model so that other applications can use the trained model entails import, development, configuration, testing, scheduling, checkpoint orchestration, training and export activities by ML tool/environment so that resulting ML model and the training pipeline thereof can be exported to other environments.
Therefore, based on possibility to deploy a model via a HTTP site (deploy 308, URL 310 - Fig. 3) it would have been obvious for one of ordinary skill in the art before the effective filing date of the invention to implement the ML orchestration environment so that a import/export capability is attached therewith so that developed code, inference calls are part of training the ML model developed via orchestration of environment so that training data can be imported for the model to be developed, tested and trained and converted into a model/pipeline so that training result all ML and pipeline inference calls, commands are packaged for export as set forth in Siracusa, whereby both ML inference code and ML pipeline software are exported outside of the training framework for use by other applications; because
providing a ML development and orchestration framework with capabilities to import data for developers to customize ML model, to train its input, to test and evaluate inference calls by the model as well as commands associated with ML pipeline orchestrated by the framework so that result from a accomplished model software and its pipeline code can be made available for export to outside host environments would ensure that model when properly validated can be re-integrated as known good models back to backend database, where possibility to export these trained model would also enhance the portability and distribution of developed software achieved by the framework, thereby boosting the marketability of the ML software being developed and garnering higher confidence by the users in the use of the developed and tested software provided as verified products or reliable learning tool by the environment.
As per claim 2, Mueller discloses computing system of claim 1, wherein:
the training dataset comprises a structured training dataset having data associated with a
number of labels (e.g. test_data_nolab = test_data_drop(labels = {LABEL-COLUNN], axis = 1 – para 0050); and
the operations further comprise receiving a selection from the user of one of the labels (manual labeling – claim 15 pg. 23; partially or un-labeled training dataset, … may engage the user … to create a job for a labeling service at circle 4C, a human labeling service – para 0047; user may also provide … data to be labeled/inferred … generate inferences for the user – para 0019) as a predicted label to be predicted by the machine learning model (predictor = task.fit(train_data, label=TARGET-COLUMN, output_directory = OUTPUT-LOCATION – para 0050)
As per claim 3, Mueller discloses computing system of claim 1, wherein the operations further comprise:
receiving a problem statement from the user, wherein the problem statement is expressed
in a natural language (user may utilize a simple UI to provide a training dataset of interest – para 0019; People_Train.csv, plaintext file CSV, TSV file – para 0037; training dataset 202, 204, 206 - Fig. 2); and
inferring, based on the problem statement (see CSV file from above; Train_csv, Test_csv, infer_csv – Fig. 2; output, inference data 111 – Fig. 3) one or more parameters of an optimization domain (e.g. target variable 402 (e.g. a value to be inferred by the desired ML model – para 0039; optimizer system 116, infer one or more probabilistic schema of the dataset, feature importance, type of the ML problem – para 0059) associated with the model architecture search (refer to claim 1) performed by the origination machine learning pipeline.
As per claim 5, Mueller discloses computing system of claim 1, wherein executing the origination machine learning pipeline to perform the model architecture search (CIVIL service …explores ML pipeline solutions – para 0021; identify a promising and diverse set of feature processing pipelines to apply to the customer dataset along with code for … feature processing model… also recommend hardware (compute instance type and count) identifiers of which hyperparameters and parallelization should use … when exploring the pipeline – para 0059) comprises:
detecting a semantic type (Train, Test, Infer – Fig. 2; “_TEST” - para 0042;”_INFER” – para 0044) for one or more features (e.g. Train_csv, Test_csv, infer_csv – Fig. 2; testing dataset – para 0042) of a plurality of features included in the training dataset; and
constraining the model architecture search (refer to claim 1) to candidate model architectures capable of processing the semantic type (Train_csv, Test_csv, infer_csv – Fig. 2) detected for the one or more features of the plurality of features included in the training dataset (Fig. 2).
As per claim 6, Mueller discloses computing system of claim 1, wherein executing the origination machine learning pipeline to perform the model architecture search (CIVIL service …explores ML pipeline solutions – para 0021; para 0059-0060) comprises:
generating one or more statistics (maximum number of training jobs, maximum runtime per job, maximum cost per exploration, number of compute instances, MSE, RMSE – para 0045) descriptive of the training dataset; and predicting one or more system settings (attempt to create an ensemble model … in some cases, an ensemble model can provide improved accuracy … that combines several base models… may learn to correct errors of other models, combines predictions from multiple other models … based on previous explorations – para 0045) of the model architecture search based on the one or more statistics (see above) descriptive of the training dataset.
As per claim 7, Mueller discloses computing system of claim 6, wherein predicting one or more system settings of the model architecture search (CIVIL service …explores ML pipeline solutions – para 0021; para 0059-0060) based on the one or more statistics descriptive of the training dataset comprises processing data descriptive of the one or more statistics (e.g. maximum number of training jobs, maximum runtime per job, maximum cost per exploration, number of compute instances, MSE, RMSE – para 0045) with a machine-learned settings (attempt to create an ensemble model … in some cases, an ensemble model can provide improved accuracy … that combines several base models… may learn to correct errors of other models, combines predictions from multiple other models … based on previous explorations – para 0045) prediction model to directly predict the system settings as an output (combines predictions from multiple other models … based on previous explorations – para 0045) of the machine-learned settings prediction model.
As per claim 8, Mueller discloses computing system of claim 1, wherein executing the origination machine learning pipeline to perform the model architecture search comprises:
generating one or more statistics (refer to claim 6) descriptive of the training dataset;
identifying, based on the one or more statistics descriptive of the training dataset, one or
more previous searches performed on one or more previous datasets (an ensemble model can provide improved accuracy … that combines several base models… may learn to correct errors of other models, combines predictions from multiple other models … based on previous explorations – para 0045) that have statistics similar to the statistics of the training dataset; and
defining one or more system settings of the model architecture search based on previous
system settings identified for the one or more previous searches (see learn to correct errors of other models, combines predictions from multiple other models, based on previous explorations – para 0045).
As per claim 9, Mueller discloses computing system of claim 1, wherein the operations further comprise:
storing metadata (metadata describing names and datatypes of the columns – para 0037; Dataset_URL, Job_name, Train_instance_type = X1.Y2.2XL; instance_type=ML.M5.Xlarge – define feature 810 - Fig. 8) descriptive of the training dataset and performance (maximum number of training jobs, maximum runtime per job, maximum cost per exploration, number of compute instances, MSE, RMSE – para 0045; recommender system 116 … metadata collected to identify a promising set of feature processing pipelines to apply to the … dataset – para 0059) of the model architecture search; and
tuning one or more parameters (prepare dataset 805, define feature processing 810, model tuning 820 – Fig. 8 and related text; recommender system 112, hyperparameter tuning – para 0064) of the origination machine learning pipeline based on the metadata descriptive (see para 0037, 0059) of the training dataset and performance (see para 0045 from above) of the model architecture search.
As per claim 10, Mueller discloses computing system of claim 1, wherein the operations further comprise identifying one or more correlations between feature crosses and labels of the training dataset (the CML service 103 may engage with the user to create a job for a labeling at circle 4C, which could be an active learning type system – para 0047 – Note3: use of labeling by the user to correlate a feature/job with a given label service reads on identifying one or more correlation between a learning feature, job with one or more user-supplied labels)
As per claim 11, Mueller discloses computing system of claim 1, wherein the deployment machine learning pipeline (Fig. 2-3) comprises both:
a fixed feature training component configured to retrain the machine learning model (e.g. if outputs from one or more additional trained machine learning models are needed … the above referenced process can be repeated with respect to the third ML scoring – para 0129) with
a fixed list of feature columns (training dataset 202 … data values arranged …one or more columns, metadata describing names and datatypes of the columns – para 0037; para 0086) ; and
a re-tuning component configured to perform a second model architecture search to
identify a new machine learning model (such hyperparameter tuning … techniques … can be utilized to work to repeatedly find better pipelines – para 0070) for the training dataset.
As per claim 12, Mueller discloses a computer-implemented method for automatic production of machine learning models and corresponding deployment pipelines, the method performed by one or more computing devices and comprising:
importing a training dataset associated with a user;
executing an origination machine learning pipeline,
wherein executing the origination machine learning pipeline comprises:
performing a model architecture search that selects and trains a machine learning model
for the training dataset; and
generating a deployment machine learning pipeline for deployment of the machine
learning model; and
exporting the machine learning model and the deployment machine learning pipeline for
deployment of the machine learning model with the deployment machine learning pipeline.
( All of which having been addressed in claim 1)
As per claim 13, refer to rejection of claim 2.
As per claim 14, refer to rejection of claim 3.
As per claim 16, refer to rejection of claim 5.
As per claim 17, refer to rejection of claim 6.
As per claim 18, refer to rejection of claim 7.
As per claim 19, refer to rejection of claim 11.
As per claim 20, Mueller discloses one or more non-transitory computer-readable media that collectively store instructions that, when executed by one or more processors of a computing system, cause the computing system to perform operations, the operations comprising:
importing a training dataset associated with a user;
executing an origination machine learning pipeline to:
perform a model architecture search that selects and trains a machine learning
model for the training dataset; and
generate a deployment machine learning pipeline for deployment of the machine
learning model; and
exporting the machine learning model and the deployment machine learning
pipeline for deployment of the machine learning model with the deployment machine learning
pipeline.
( All of which having been addressed in claim 1)
As per claim 21, Mueller discloses computing system of claim 1, wherein the operations further comprise:
receiving a request for an operational machine learning model trained using the training
dataset associated with the user (serverless function can be invoked through … HTTP request message. Accordingly, users can define serverless functions that can be executed on-demand … the CML service makes the ML model building process easier … user does not need to write any commands, function calls – para 0028-0029; user can use it … issuing inference request – para 0055; receive inference requests from outside clients – para 0073; user can provide training request – para 0094, 0097; identify … training data indicated by the training request and select a machine learning model to train – para 0101), wherein the origination machine learning pipeline is executed (refer to perform a model search, generating a deployment in claim 1) based on the request.
As per claim 22, Mueller discloses computing system of claim 21, wherein executing the origination machine learning pipeline further comprises:
evaluating a performance quality (quality metrics … level of confidence that the accuracy of the machine learning model being trained is known, evaluator … compare the outputs and determine quality metrics … based on comparison … distance between the machine learning model outputs and the expected outputs – para 0112) of the machine learning model after training the
machine learning model with the training dataset; and
validating the machine learning model satisfies the request for the operational machine
learning model based on the performance quality (see para 0112; optimizer system … identify additional … ML variants to test that are likely to result in a good model … such … tuning techniques … can be utilized to … repeatedly find better pipelines – para 0070) of the machine learning model.
Conclusion
Any inquiry concerning this communication or earlier communications from the examiner should be directed to Tuan A Vu whose telephone number is (571) 272-3735. The examiner can normally be reached on 8AM-4:30PM/Mon-Fri.
If attempts to reach the examiner by telephone are unsuccessful, the examiner's supervisor, Chat Do can be reached on (571)272-3721.
The fax phone number for the organization where this application or proceeding is assigned is (571) 273-3735 ( for non-official correspondence - please consult Examiner before using) or 571-273-8300 ( for official correspondence) or redirected to customer service at 571-272-3609.
Any inquiry of a general nature or relating to the status of this application should be directed to the TC 2100 Group receptionist: 571-272-2100.
/Tuan A Vu/
Primary Examiner, Art Unit 2193
August 26, 2026
