Builder project·10-15 hours

End-to-End Tabular ML Capstone

Independently take a licensed tabular dataset from question and data audit through a leakage-safe pipeline, error analysis, model card, and bounded deployment recommendation.

Scenario

Choose a small public dataset with clear provenance and a defensible prediction time. Act as both builder and reviewer: create reproducible evidence, challenge your own conclusions, and make deployment optional rather than assumed.

You will demonstrate

Complete an independent reproducible ML workflow.
Compare a baseline and at least two justified candidates on identical evidence.
Communicate errors, limitations, and an operating decision through a portfolio artifact.

Project evidence

Show the work, not a checked box.

Each response is stored in this browser as you type. Include metrics, test output, or a decision rationale wherever the deliverable asks for it.

Problem and data contract

Define user, row, target, prediction time, feature availability, error costs, source license, and prohibited uses.

Required evidence: A reproducible problem statement, data dictionary, provenance record, and go/no-go data audit.

Your evidence

0/80 minimum characters

Reproducible pipeline

Create immutable splits and keep imputation, encoding, scaling, feature selection, and fitting inside evaluated pipelines.

Required evidence: Versioned code, environment details, split rationale, and passing data/pipeline tests.

Your evidence

0/80 minimum characters

Baseline and model comparison

Compare a simple baseline with at least two candidates using identical folds and decision-aware metrics.

Required evidence: Fold-level metric table, uncertainty summary, selected threshold, and complexity rationale.

Your evidence

0/80 minimum characters

Error and stress analysis

Inspect concrete errors, meaningful slices, missingness, perturbations, and one plausible shift scenario.

Required evidence: Error taxonomy, slice table with sample sizes, and one controlled stress result.

Your evidence

0/80 minimum characters

Model card and decision memo

Document intended use, non-goals, evaluation, limitations, ethical considerations, monitoring, fallback, and rollback.

Required evidence: A model card plus launch, revise, narrow, or stop recommendation tied to evidence.

Your evidence

0/80 minimum characters

Runnable Python lab

Select from fold-level evidence

Summarize cross-validation scores and reject a candidate that fails the stated minimum or stability gate.

Test output

Run the code when you are ready. The first load downloads the in-browser Python runtime.

Code runs only in a time-limited browser worker. It is not sent to the IntuitiveAI server. Do not paste secrets.

Expected result

linear mean 0.79 with low fold variation
selected: linear
ALL TESTS PASSED

Project defense

Your most complex model improves mean cross-validation F1 by 0.004 but doubles latency and fold variation overlaps heavily. What is the strongest conclusion?

Rubric self-review

Rate the evidence, not your effort: 0 missing, 1 weak, 2 adequate, 3 strong. All criteria must be reviewed, but a low honest score does not get hidden.

Data provenance, license, target, prediction time, and split boundaries are explicit.

The runnable pipeline prevents train/evaluation contamination.

Comparisons use identical folds and include a meaningful baseline.

Error and slice analysis support concrete, appropriately bounded conclusions.

The portfolio artifact is reproducible and the decision includes monitoring and rollback.

Artifact

Attach the portfolio evidence.

Submit a complete .ipynb, a Markdown report with repository link, or a PDF portfolio report. The platform never executes uploaded notebooks.

Validate locallyUpload privately

Anonymous mode stores only metadata and a fingerprint locally. Private upload requires an account and an explicit file selection. Uploaded notebooks are never executed.

Useful references

Model Cards for Model ReportingA reporting framework for intended use, evaluation context, limitations, and ethical considerations.The ML Test ScoreA practical rubric for testing and monitoring production ML systems.UCI Machine Learning RepositoryPublic datasets with source and task documentation; verify each dataset license before use.

Project completion gate

Completion is controlled by stored evidence, deterministic tests, a decision defense, rubric review, and the artifact when required.

Evidence pendingCode pendingDefense pendingRubric pendingArtifact pending

Optional cloud portfolio

Submit evidence across devices.

An account is required. Submit only when the local completion gate passes. AI review is advisory and separate from deterministic completion.

After submission, I may choose to send the stored evidence text to the configured AI model for advisory rubric feedback. Artifact bytes are not sent by this review action.

Account settings