Add example MLOps job scripts for Databricks, SageMaker and Snowflake by Vedaant-J · Pull Request #71 · NVIDIA/kumo-engine

Vedaant-J · 2026-10-21T01:00:44Z

Series: kumo-engine MLOps series: every snapshot, training table, model and prediction run carries a self-contained _provenance.json; with kumo.mlops(...) stores each artifact under one artifact root, tracks fits and prediction runs in MLflow or Snowflake and registers models; volumes and stages can hold files; prediction runs are monitored for drift and realized performance.
Stack: Models, registry and job templates: #57 → #59 → #65 → #71
Merge order: one linear stack; merge bottom to top (#53 first)
This PR: Adds job templates that run kumo-engine on Databricks Jobs, SageMaker Pipelines, and Snowflake ML Jobs with Tasks.

Summary

This PR adds job templates that run kumo-engine on Databricks Jobs, SageMaker Pipelines, and Snowflake ML Jobs with Tasks. Two plain Python job scripts take CLI flags and run inside with kumo.mlops(kumo.MLOps(...)), so every artifact lands under one artifact root. Per-platform wiring schedules them, passes the flags, tracks the fit, and hands the trained model to the scoring job: by registry alias, or by its models/<bundle id> path without a registry.

Problem. A team that wants to retrain weekly and score daily can't answer these questions today:

How do I run fit and predict as a Databricks job, a SageMaker pipeline step, or a Snowflake Task?
How do I turn MLOps on in a scheduled script, so the snapshot, training table and model of each run are stored and traceable?
Where does the scoring job find the model the training job produced: an MLflow alias, a Snowflake registry alias, or a path?

Today none of these can be answered without reading engine source.

Cause. The repository ships only a local SDK example:

examples/
└── sdk-local/
    ├── README.md
    └── check.py

Change.

examples/
├── sdk-local/
└── mlops/                                   # new
    ├── README.md        # overview, flags, local run
    ├── fit_job.py       # inside kumo.mlops(): snapshot, training table, fit (stored, tracked, registered)
    ├── score_job.py     # load "champion" or --model-path, fresh or stored snapshot, trainer.predict
    ├── jobs.py          # flags, graph (users, items, orders), connectors, kumo.MLOps settings
    ├── databricks/  databricks.yml  README.md
    ├── sagemaker/   pipeline.py  README.md
    └── snowflake/   register_jobs.py  README.md

Platform	Train	Score	Model handoff
Databricks (Asset Bundle)	`kumo_train`: `fit_job.py` on a single-node CPU job cluster	`kumo_score`: daily `score_job.py` into a UC table (created paused)	`--registry mlflow`: `MLOps(tracking=MlflowTracking(), registry=ModelRegistryLocation("<catalog>.<schema>.users_churn", alias="champion"))` (#65); scoring loads `models:/…@champion`
SageMaker Pipelines (SDK v2)	`Fit` training step	`Predict` processing step	`--registry mlflow` against SageMaker managed MLflow; scoring loads `models:/users-churn@champion`
Snowflake ML Jobs + Tasks (experimental: preview APIs, not run)	`KUMO_TRAIN`: weekly	`KUMO_SCORE`: daily into a table	`--registry snowflake`: `MLOps(tracking=SnowflakeTracking(...), registry=ModelOutputConfig(sf, location=ModelRegistryLocation("USERS_CHURN", alias="champion")))` (#59); scoring loads `champion`
Plain VM / workstation	`python fit_job.py`	`python score_job.py`	no registry: the fit prints `model: models/<bundle id>`; scoring loads `--model-path models/<bundle id>` from the artifact root

The scripts use only the public API. Each builds a kumo.MLOps from its flags and runs inside with kumo.mlops(cfg). The snapshot, the training table and the model are stored under --artifact-root when they are produced (Export and load graph snapshots through any file destination #54, Record training-table lineage across export and reload #55, Record bundle_id and fit provenance in every model bundle #56). The fit is tracked and the model registered when --registry is given. Scoring writes --predictions-table when given.
Flags, no environment. --artifact-root, --source (databricks, snowflake or a Parquet directory), --catalog/--schema/--warehouse, --registry, --model-name, --model-path, --snapshot, --predictions-table, --query, --num-experiments and --scratch-dir. Credentials come from the platform identity. Only MLflow's own settings (MLFLOW_TRACKING_URI, ...) stay environment variables.
No job ids, no summary line. The scripts exit 0 on success and non-zero with a traceback on failure. The scheduler tracks its own runs.
Every score passes a binary threshold: 0.5, unless the model carries Add TrainingResult.with_scoring_defaults and Trainer.export #57 scoring defaults (trainer.result.scoring_defaults). The template query (PREDICT COUNT(orders.*, 0, 30, days) > 0 FOR EACH users.USER_ID) is binary, and a binary predict without a threshold fails.
Databricks.
- Each task is a spark_python_task whose parameters are the flags. The artifact root is /Volumes/<catalog>/<schema>/<volume>/kumo, and the cluster env sets MLFLOW_TRACKING_URI=databricks and MLFLOW_REGISTRY_URI=databricks-uc.
- Each task installs a requirements file, then the kumo-engine wheel. The file starts with --prefer-binary, because DBR 16.4's setuptools cannot build the SQLAlchemy sdist MLflow pulls.
- A spark_python_task runs the script without __file__, so the scripts find their directory from sys.argv[0].
SageMaker. The training step passes the flags as container_arguments and the processing step as arguments. MLFLOW_TRACKING_URI is the tracking-server or MLflow App ARN, through sagemaker-mlflow. --instance-type is required.
Snowflake (experimental; built on preview APIs and not run):
- MLJobDefinition.register defines the train and score jobs. Each Task runs MLJobDefinition.to_sql(job_args=<flags>, use_async=False) directly; the old task.sql existed only to pass job ids.
- Tracking and the registry use the connector's own Snowpark session through the new SnowflakeConnector.session().
- The artifact root must be an s3:// or gs:// URI that the external access integration reaches.

Real output of fit_job.py then score_job.py on a workstation, CPU, no registry. Data: --source <data dir> with synthetic users/items/orders (50 users, 20 items, 3,000 orders over 120 days), a 7-day query, one experiment. Hashes shortened:

$ python fit_job.py --source <data dir> --artifact-root <root> --query 'PREDICT COUNT(orders.*, 0, 7, days) > 0 FOR EACH users.USER_ID' --num-experiments 1
model: models/f48e17f0…
$ python score_job.py --source <data dir> --artifact-root <root> --model-path models/f48e17f0…
predictions: 50 rows

<root>/
  snapshots/7ff1fc83…/bee34f9aa0ea3e83/      _provenance.json  kumo_snapshot.json  index/…
  training_tables/tt-20b4fc72…/              _provenance.json  _SUCCESS  part-00000.parquet …
  models/f48e17f0…/                          _provenance.json  metrics.json  publication-manifest.json  model/  manifests/ …

Back-compat

Examples plus one additive connector method.

Situation	Behaviour
Existing library calls	Unchanged.
`SnowflakeConnector.session()`	New; returns the Snowpark session the connector already reads through.
Engine wheel (CI runs tests from it)	`test_job_templates.py` skips when `examples/mlops` is absent.

Known limitations

An earlier revision of the Databricks bundle and the SageMaker pipeline, configured through environment variables with the same engine calls, was deployed and run live. Both jobs/steps exited 0, the fit registered and moved champion, and scoring loaded it. This flag-based revision is verified locally only. The Snowflake template has not been run and is marked experimental.
Without --predictions-table, score_job.py stores nothing yet; the prediction-run store under predictions/<id>/ arrives with Record a manifest and output score statistics for every prediction run #67.
Use a standard (non-ML) Databricks runtime: the ML runtime's preinstalled torch, torchvision and pyOpenSSL conflict with kumo-engine's pins. Databricks GPU nodes need a GPU ML runtime, so the bundle uses a CPU node type with local NVMe disks.
SYSTEM$EXECUTE_ML_JOB, which MLJobDefinition.to_sql generates, is not a documented public API.

Testing Done

test_job_templates.py:
- fit_job.py then score_job.py end to end on local Parquet (no registry): the fit prints model: models/<id>, the root holds models/<id>/_provenance.json, snapshots/… and training_tables/…, and the score loads --model-path models/<id> and scores every user;
- the Databricks tasks pass the same --artifact-root and --registry mlflow to both scripts, and --predictions-table to the score job only;
- the Snowflake tasks run each job's to_sql statement with the flags (--predictions-table on the score job only);
- both scripts import when executed without __file__ (Databricks spark_python_task);
- the SageMaker pipeline refuses to run without --instance-type, and SageMaker and Snowflake name the scripts;
- every Python file under examples/mlops byte-compiles.
test_snowflake_connector.py: session() is the reader's session for the connector's target.
On this head, test_job_templates.py + test_snowflake_connector.py gave 20 passed. flake8 is clean on examples/mlops.

Checklist

No credentials or real identifiers; placeholders only
Tests added

🤖 Generated with Claude Code

Vedaant-J · 2026-10-21T01:00:46Z

Change map

File	Function / scope	What changed	Why
`examples/mlops/README.md`	overview	new: the two job scripts inside `kumo.mlops()`, platform table (Snowflake marked experimental), flag table, local run with `--model-path models/<bundle id>`, exit status	one index the platform READMEs point to
`examples/mlops/databricks/README.md`	docs	new: jobs, prerequisites (requirements file with `--prefer-binary` for DBR 16.4, experiment parent folder, `dev` target), privileges, `__file__` fallback, what was run live	Databricks setup
`examples/mlops/databricks/databricks.yml`	`kumo_train`, `kumo_score`	new: `spark_python_task` per script with the flags as `parameters` (volume artifact root, `--registry mlflow`), MLflow env for Unity Catalog, daily schedule (paused), `requirements` library before the wheel, default `dev` target	Databricks Asset Bundle
`examples/mlops/fit_job.py`	`main`	new: inside `kumo.mlops(jobs.mlops(...))`: snapshot, training table, fit (stored, tracked, registered by the MLOps settings); prints `model: models/<id>`; finds its directory without `__file__`	the training job every platform runs
`examples/mlops/jobs.py`	`parse_args`, `source`, `mlops`, `build_graph`, `load_model`, `prediction_output`	new: CLI flags; connectors; `kumo.MLOps` with `MlflowTracking` + `ModelRegistryLocation`, `SnowflakeTracking` + `ModelOutputConfig`, or root only; load by alias or `--model-path`; optional predictions table	one shared module the scripts and platforms configure
`examples/mlops/sagemaker/README.md`	docs	new: image and code channel, IAM, MLflow ARN and client/server versions, v2 → v3 mapping, what was run live	SageMaker setup
`examples/mlops/sagemaker/pipeline.py`	`build_pipeline`, `main`	new: training step runs `fit_job.py`, processing step `score_job.py`, both with the flags; `--instance-type` required	SageMaker Pipelines
`examples/mlops/score_job.py`	`main`	new: inside `kumo.mlops()`: load `champion` or `--model-path`, fresh or stored snapshot, `trainer.predict` with a threshold unless the model has scoring defaults; finds its directory without `__file__`	the scoring job every platform runs
`examples/mlops/snowflake/README.md`	docs	new: marked experimental (preview APIs); jobs and tasks, alias promotion, tracking session, privileges, unverified items	Snowflake setup
`examples/mlops/snowflake/register_jobs.py`	`register`, `task_sql`, `main`	new (experimental): ML Job definitions with the scripts as entrypoints; each Task runs `to_sql(job_args=<flags>)`	Snowflake ML Jobs
`packages/kumo-engine/kumo/engine/connector/snowflake.py`	`SnowflakeConnector.session`	new: the Snowpark session the connector reads through	tracking and registry reuse the reader's session in a job container
`packages/kumo-engine/test/test_job_templates.py`	new tests	scripts compile; Databricks and Snowflake tasks pass the flags; fit then score end to end on local Parquet with the artifact-root layout and `--model-path`; scripts import without `__file__`; `--instance-type` required	templates stay valid as the API changes
`packages/kumo-engine/test/test_snowflake_connector.py`	`test_session_is_the_readers_session`	`session()` comes from the reader's session cache for the connector's target	coverage

Requirement: show how Databricks, SageMaker and Snowflake schedule kumo-engine jobs with MLOps on, track the fit and hand the model to scoring. Existing: kumo.MLOps (#50), the artifact-root stores (#54–#56), MLflow and Snowflake tracking and registries (#59, #60, #65, #68). Smallest: two plain scripts, one shared module, per-platform wiring, one connector method. Stopping condition: fit then score run end to end locally from the scripts, and each README states what was run live and what was not.

copy-pr-bot · 2026-10-21T01:00:48Z

This pull request requires additional validation before any workflows can run on NVIDIA's runners.

Pull request vetters can view their responsibilities here.

Contributors can view more details about this message here.

coderabbitai · 2026-10-21T01:00:52Z

Warning

Review limit reached

Enable usage-based reviews in Billing to review now. Otherwise, wait until the next included review is available.

Next included review available in 24 minutes.

Check out review usage here.

View limit details

Limit details: You’ve used all 12 included reviews currently available.

Learn how review limits work.

Review configuration:

Run configuration

Source mode: Auto
Repository: NVIDIA/kumo-engine
Branch: vedaant/mlops-mlflow-model
File: .coderabbit.yaml
Commit: a8aad9a
Layers: Repository YAML
Review profile: QUIET
Plan: Enterprise
Run ID: d1c6a8f5-6ed5-4c39-a909-aa4dc2344316

Commits

Reviewing files that changed from the base of the PR and between ecf304a and 1ea72bb.

Files selected for processing: 13

examples/mlops/README.md
examples/mlops/databricks/README.md
examples/mlops/databricks/databricks.yml
examples/mlops/fit_job.py
examples/mlops/jobs.py
examples/mlops/sagemaker/README.md
examples/mlops/sagemaker/pipeline.py
examples/mlops/score_job.py
examples/mlops/snowflake/README.md
examples/mlops/snowflake/register_jobs.py
packages/kumo-engine/kumo/engine/connector/snowflake.py
packages/kumo-engine/test/test_job_templates.py
packages/kumo-engine/test/test_snowflake_connector.py

Walkthrough

Adds shared training and scoring scripts with options for source data, model registration or loading, snapshots, and prediction output. Adds Databricks, SageMaker, and Snowflake configurations for running these jobs. The SageMaker pipeline runs a training step before its dependent scoring step. Tests cover job configuration, script loading, local fit-then-score execution, and Snowflake session access.

Sequence Diagram(s)

sequenceDiagram
  participant SageMakerPipeline
  participant fit_job.py
  participant MLflowRegistry
  participant score_job.py
  SageMakerPipeline->>fit_job.py: Run training step
  fit_job.py->>MLflowRegistry: Register model and update champion alias
  SageMakerPipeline->>score_job.py: Run dependent scoring step
  score_job.py->>MLflowRegistry: Load champion model

Priority: Normal

Merge Risk: Moderate · up to ecf30

The SageMaker example pipeline runs scoring but keeps no predictions, so a successful run produces nothing usable. Add a durable predictions destination, or document the limitation, before merging. The Databricks and Snowflake templates and the shared scripts are otherwise sound as an example.

Pre-merge checks: 2 · 1

Failed checks: 1 inconclusive

Check name	Status	Explanation	Resolution
Change Map Design Review		The required PR conversation comment beginning with `<!-- kumo-pr-change-map -->` is not available. The supplied review context contains no such comment, and the pull-request description is explicitly…	Provide the complete change-map conversation comment and the omitted portion of the PR description, or make the linked conversation data available. Then review each mapped change against the actual diff and repository behavior.

Passed checks: 2 passed

Check name	Status	Explanation
Pr Size		The pull request changes 13 files. The failure threshold is more than 40 files, so this check passes.
Public Resource Access		The pull request adds public documentation URLs for Databricks, AWS SageMaker, and Snowflake. Anonymous HEAD requests returned HTTP 200 for all 14 added URLs. Other URI-like values are local paths or …

Full details: Public Resource Access

Explanation

The pull request adds public documentation URLs for Databricks, AWS SageMaker, and Snowflake. Anonymous HEAD requests returned HTTP 200 for all 14 added URLs. Other URI-like values are local paths or illustrative placeholders, which this check excludes. No added external resource reference is confirmed inaccessible.

Full details: Change Map Design Review

Explanation

The required PR conversation comment beginning with  is not available. The supplied review context contains no such comment, and the pull-request description is explicitly truncated. The repository only contains instructions describing that comment; it does not contain the comment itself. Therefore, this check cannot assess the proposed design against its stated rationale, smallest sufficient change, verification, and stopping condition.

Resolution

Provide the complete change-map conversation comment and the omitted portion of the PR description, or make the linked conversation data available. Then review each mapped change against the actual diff and repository behavior.

Finishing Touches: 💡 1

🛠️ Fix failing CI checks: 💡

Commit to this branch
Create a new PR

Autofix · Keep fixing CodeRabbit findings and required CI, and resolving merge conflicts

_{Comment @coderabbitai help to get the list of available commands.}

coderabbitai

Actionable comments posted: 1

Review info

Run configuration

Source mode: Auto
Repository: NVIDIA/kumo-engine
Branch: vedaant/mlops-mlflow-model
File: .coderabbit.yaml
Commit: 9b8a83a
Layers: Repository YAML
Review profile: QUIET
Plan: Enterprise
Run ID: fa4abf28-62b5-4f3c-95ba-3c9df2d4ea74

Commits

Reviewing files that changed from the base of the PR and between 9b8a83a and ecf304a.

Files selected for processing: 13

examples/mlops/README.md
examples/mlops/databricks/README.md
examples/mlops/databricks/databricks.yml
examples/mlops/fit_job.py
examples/mlops/jobs.py
examples/mlops/sagemaker/README.md
examples/mlops/sagemaker/pipeline.py
examples/mlops/score_job.py
examples/mlops/snowflake/README.md
examples/mlops/snowflake/register_jobs.py
packages/kumo-engine/kumo/engine/connector/snowflake.py
packages/kumo-engine/test/test_job_templates.py
packages/kumo-engine/test/test_snowflake_connector.py

Included review availability: This review used your included allowance. Your plan provides up to 12 included reviews per hour; 0 remain after this review.

coderabbitai · 2026-10-24T19:20:40Z

+            arguments=[
+                *flags, '--scratch-dir', '/opt/ml/processing/scratch'
+            ],


Major · Data Integrity & Integration · Quick win

Persist the processing step’s predictions.

Predict passes no --predictions-table or other output destination. score_job.main() therefore calls trainer.predict() without an output configuration, prints only a row count, and exits. The pipeline can report success without retaining any predictions. Add a supported persistent output destination to this step. (docs.aws.amazon.com)

Teams want to run kumo-engine fits and scoring on the scheduler they already have, but there is no runnable example of wiring the engine, MLOps settings and a registry into a scheduled job. examples/mlops holds two plain Python job scripts (fit_job.py, score_job.py) and a shared jobs.py. They take CLI flags (--artifact-root, --source, --registry, --model-name, --model-path, --predictions-table, ...), build kumo.MLOps from them and run inside `with kumo.mlops(cfg)`, so snapshots, training tables and models are stored under the artifact root (models/<bundle id>/). With a registry the fit registers the model and moves `champion`; scoring loads it, or --model-path from the root. Platform wiring: a Databricks Asset Bundle, a SageMaker Pipeline and (experimental) Snowflake ML Jobs + Tasks, each passing the flags. SnowflakeConnector.session() exposes the connector's Snowpark session for SnowflakeTracking. Verified with test_job_templates.py (local fit then score) and test_snowflake_connector.py. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

Vedaant-J · 2026-10-26T18:31:54Z

Folded into #64 (prediction runs & monitoring): the job scripts are now examples under examples/mlops in #64, and SnowflakeConnector.session() moved to #60. The consolidated stack shows only the end state.

Vedaant-J temporarily deployed to gha-kumo-workflow-tests October 21, 2026 01:00 — with GitHub Actions Inactive

Vedaant-J force-pushed the vedaant/mlops-cli branch from 65608c7 to 7014f45 Compare October 21, 2026 01:56

Vedaant-J force-pushed the vedaant/mlops-job-templates branch from 91d11cd to 29f8c7f Compare October 21, 2026 01:56

Vedaant-J temporarily deployed to gha-kumo-workflow-tests October 21, 2026 01:56 — with GitHub Actions Inactive

Vedaant-J force-pushed the vedaant/mlops-cli branch from 7014f45 to 7a38c4b Compare October 21, 2026 07:18

Vedaant-J force-pushed the vedaant/mlops-job-templates branch from 29f8c7f to 2aa4786 Compare October 21, 2026 07:18

Vedaant-J temporarily deployed to gha-kumo-workflow-tests October 21, 2026 07:18 — with GitHub Actions Inactive

Vedaant-J force-pushed the vedaant/mlops-cli branch from 7a38c4b to 6b20cb6 Compare October 21, 2026 07:57

Vedaant-J force-pushed the vedaant/mlops-job-templates branch from 2aa4786 to 937102d Compare October 21, 2026 07:57

Vedaant-J temporarily deployed to gha-kumo-workflow-tests October 21, 2026 07:57 — with GitHub Actions Inactive

Vedaant-J force-pushed the vedaant/mlops-cli branch from 6b20cb6 to 10c4a23 Compare October 21, 2026 08:58

Vedaant-J force-pushed the vedaant/mlops-job-templates branch from 937102d to dd4e481 Compare October 21, 2026 08:59

Vedaant-J temporarily deployed to gha-kumo-workflow-tests October 21, 2026 08:59 — with GitHub Actions Inactive

Vedaant-J force-pushed the vedaant/mlops-cli branch from 10c4a23 to 573186e Compare October 21, 2026 09:02

Vedaant-J force-pushed the vedaant/mlops-job-templates branch from dd4e481 to 7b51c15 Compare October 21, 2026 09:02

Vedaant-J temporarily deployed to gha-kumo-workflow-tests October 21, 2026 09:02 — with GitHub Actions Inactive

Vedaant-J mentioned this pull request Oct 21, 2026

Add kumo.job() for scheduled job scripts #70

Closed

3 tasks done

Vedaant-J force-pushed the vedaant/mlops-cli branch from 573186e to 62b64b6 Compare October 21, 2026 10:06

Vedaant-J force-pushed the vedaant/mlops-job-templates branch from 7b51c15 to cb68ce2 Compare October 21, 2026 10:06

Vedaant-J temporarily deployed to gha-kumo-workflow-tests October 21, 2026 10:06 — with GitHub Actions Inactive

Vedaant-J force-pushed the vedaant/mlops-cli branch from 62b64b6 to 04f3ea1 Compare October 21, 2026 16:10

Vedaant-J force-pushed the vedaant/mlops-job-templates branch from cb68ce2 to e203171 Compare October 21, 2026 16:10

Vedaant-J temporarily deployed to gha-kumo-workflow-tests October 21, 2026 16:11 — with GitHub Actions Inactive

Vedaant-J force-pushed the vedaant/mlops-cli branch from 04f3ea1 to 913d621 Compare October 21, 2026 16:41

Vedaant-J force-pushed the vedaant/mlops-job-templates branch from e203171 to af92e85 Compare October 21, 2026 16:42

Vedaant-J had a problem deploying to gha-kumo-workflow-tests October 21, 2026 16:42 — with GitHub Actions Error

Vedaant-J temporarily deployed to gha-kumo-workflow-tests October 21, 2026 16:44 — with GitHub Actions Inactive

Vedaant-J temporarily deployed to gha-kumo-workflow-tests October 22, 2026 00:07 — with GitHub Actions Inactive

Vedaant-J force-pushed the vedaant/mlops-cli branch from 6bfa01e to 19bd626 Compare October 22, 2026 00:21

Vedaant-J force-pushed the vedaant/mlops-job-templates branch from 56b8448 to bb2b95a Compare October 22, 2026 00:21

Vedaant-J temporarily deployed to gha-kumo-workflow-tests October 22, 2026 00:21 — with GitHub Actions Inactive

Vedaant-J force-pushed the vedaant/mlops-cli branch from 19bd626 to d8a7093 Compare October 22, 2026 00:52

Vedaant-J force-pushed the vedaant/mlops-job-templates branch from bb2b95a to 57045b8 Compare October 22, 2026 00:53

Vedaant-J temporarily deployed to gha-kumo-workflow-tests October 22, 2026 00:53 — with GitHub Actions Inactive

Vedaant-J force-pushed the vedaant/mlops-cli branch from d8a7093 to caa5017 Compare October 22, 2026 02:21

Vedaant-J force-pushed the vedaant/mlops-job-templates branch from 57045b8 to da8b13f Compare October 22, 2026 02:22

Vedaant-J temporarily deployed to gha-kumo-workflow-tests October 22, 2026 02:22 — with GitHub Actions Inactive

Vedaant-J force-pushed the vedaant/mlops-job-templates branch from da8b13f to 3fe4c18 Compare October 22, 2026 04:59

Vedaant-J temporarily deployed to gha-kumo-workflow-tests October 22, 2026 04:59 — with GitHub Actions Inactive

Vedaant-J temporarily deployed to gha-kumo-workflow-tests October 22, 2026 05:29 — with GitHub Actions Inactive

Vedaant-J temporarily deployed to gha-kumo-workflow-tests October 22, 2026 05:31 — with GitHub Actions Inactive

Vedaant-J force-pushed the vedaant/mlops-cli branch from e07a332 to b3dea5a Compare October 22, 2026 05:39

Vedaant-J force-pushed the vedaant/mlops-job-templates branch from 2bfe866 to f380a89 Compare October 22, 2026 05:40

Vedaant-J temporarily deployed to gha-kumo-workflow-tests October 22, 2026 05:40 — with GitHub Actions Inactive

Vedaant-J force-pushed the vedaant/mlops-cli branch from b3dea5a to 2c18051 Compare October 22, 2026 19:16

Vedaant-J force-pushed the vedaant/mlops-job-templates branch from f380a89 to f21cf92 Compare October 22, 2026 19:16

Vedaant-J temporarily deployed to gha-kumo-workflow-tests October 22, 2026 19:16 — with GitHub Actions Inactive

This was referenced Oct 24, 2026

Add TrainingResult.with_scoring_defaults and Trainer.export #57

Closed

Add model registry aliases and a bundle id pointer on registered versions #59

Closed

Log, register and serve fitted models through MLflow pyfunc #65

Closed

Vedaant-J force-pushed the vedaant/mlops-job-templates branch from f21cf92 to f258bb5 Compare October 24, 2026 18:40

Vedaant-J changed the base branch from vedaant/mlops-cli to vedaant/mlops-mlflow-model October 24, 2026 18:40

This was referenced Oct 24, 2026

Record prediction runs and monitor score drift #64

Open

Record a manifest and output score statistics for every prediction run #67

Closed

coderabbitai Bot reviewed Oct 24, 2026

View reviewed changes

Navigation Menu

Repository navigation

Add example MLOps job scripts for Databricks, SageMaker and Snowflake - #71

Add example MLOps job scripts for Databricks, SageMaker and Snowflake#71
Vedaant-J wants to merge 1 commit into
vedaant/mlops-mlflow-modelfrom
vedaant/mlops-job-templates

Vedaant-J commented Oct 21, 2026 •

edited

Loading

Uh oh!

Vedaant-J commented Oct 21, 2026 •

edited

Loading

Uh oh!

copy-pr-bot Bot commented Oct 21, 2026

Uh oh!

coderabbitai Bot commented Oct 21, 2026 •

edited

Loading

Review limit reached

Sequence Diagram(s)

Uh oh!

coderabbitai Bot left a comment

Uh oh!

coderabbitai Bot Oct 24, 2026

Uh oh!

Vedaant-J commented Oct 26, 2026

Uh oh!

This branch was previously deployed

Reviewers

Uh oh!

Assignees

Labels

Projects

Milestone

Development

Uh oh!

1 participant

Conversation

Vedaant-J commented Oct 21, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Summary

Back-compat

Known limitations

Testing Done

Checklist

Uh oh!

Vedaant-J commented Oct 21, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Change map

Uh oh!

copy-pr-bot Bot commented Oct 21, 2026

Uh oh!

coderabbitai Bot commented Oct 21, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Review limit reached

Sequence Diagram(s)

Uh oh!

coderabbitai Bot left a comment

Choose a reason for hiding this comment

Uh oh!

coderabbitai Bot Oct 24, 2026

Choose a reason for hiding this comment

Uh oh!

Vedaant-J commented Oct 26, 2026

Uh oh!

This branch was previously deployed

Reviewers

Uh oh!

Assignees

Labels

Projects

Milestone

Development

Uh oh!

1 participant

Vedaant-J commented Oct 21, 2026 •

edited

Loading

Vedaant-J commented Oct 21, 2026 •

edited

Loading

coderabbitai Bot commented Oct 21, 2026 •

edited

Loading