Skip to content

FAQ

Do I need all projects to run?

No. Start with one project and expand gradually.

Do I need Kaggle credentials?

Only for causalml-kaggle-showcase.

Do I need cloud services for new projects?

No. All new showcase projects are laptop-first. Some projects have optional advanced extras.

Why are artifact folders mostly empty?

Some large generated artifacts are excluded from git to keep the repo lightweight. Core example artifacts and manifest contracts are kept for reproducibility.

What does make check do at root?

It runs lint, type checks, tests, and supervised contract checks across projects.

What does make check-contracts do?

It bootstraps missing supervised artifacts in quick mode and then validates required train/val/test split manifests, EDA outputs, leakage reports, and experiment logs.

Do supervised showcases always enforce train/val/test instead of train/test?

Yes. Supervised showcase contracts enforce explicit train_rows, val_rows, and test_rows in split manifests.

Where can I see which project covers each advanced topic?

Use docs/aspect-coverage-matrix.md for a direct mapping of topics to commands and output artifacts.

Is there a generated docs site for this repo?

Yes. This repo uses MkDocs Material.

  • Run locally: make docs-serve
  • Build static site: make docs-build
  • Run strict docs validation: make docs-check
  • Public URL (after Pages enablement): https://conqueror.github.io/mcgill-showcases/

Do we host API docs too?

Yes. API reference pages are part of the MkDocs site:

  • docs/api/ranking-api.md
  • docs/api/demand-api.md

Their schemas are sourced from: - docs/api/assets/openapi/ranking-api.json - docs/api/assets/openapi/demand-api.json

Why are there project-level Makefiles and a root Makefile?

  • Root Makefile: orchestrates all projects.
  • Project Makefile: detailed workflows specific to that project.

Which project should I do after supervised basics?

Choose by goal: - Production focus: mlops-drift-production-showcase. - Responsible AI focus: xai-fairness-audit-showcase. - Optimization focus: automl-hpo-showcase.

Where is the ranking workflow?

Use this two-project sequence: - projects/learning-to-rank-foundations-showcase for grouped ranking model training and NDCG evaluation. - projects/ranking-api-productization-showcase for FastAPI serving, schema contracts, structured logs, and OpenAPI export.

Where is the forecasting and observability workflow?

Use this two-project sequence: - projects/nyc-demand-forecasting-foundations-showcase for time-ordered demand forecasting with train/val/test. - projects/demand-api-observability-showcase for demand serving APIs with Prometheus metrics and optional OTel instrumentation.

Is there a direct showcase architecture map?

Yes. See docs/showcase-architecture.md for track-level mapping across projects.

Where is the contributor checklist for adding a new showcase?

Use docs/new-showcase-playbook.md. It defines required project structure, artifact contracts, tests, CI integration, and documentation updates.