Skip to content

Getting Started

Prerequisites

  • Python 3.11+
  • uv installed
  • Git

5-Minute Setup

git clone git@github.com:conqueror/mcgill-showcases.git
cd mcgill-showcases
make -C projects/sota-supervised-learning-showcase sync

Pick Your First Project

  • Deep learning math foundations: projects/deep-learning-math-foundations-showcase
  • Neural network mechanics and decision boundaries: projects/neural-network-foundations-showcase
  • PyTorch training loops and regularization: projects/pytorch-training-regularization-showcase
  • Easiest start: projects/sota-supervised-learning-showcase
  • Causal decisioning: projects/causalml-kaggle-showcase
  • Modern unlabeled-data workflows: projects/sota-unsupervised-semisup-showcase
  • Production monitoring and serving: projects/mlops-drift-production-showcase
  • Credit-risk capstone from course notebooks: projects/credit-risk-classification-capstone-showcase
  • Responsible AI auditing: projects/xai-fairness-audit-showcase
  • Rollout and rollback simulation: projects/model-release-rollout-showcase
  • Data diagnostics and leakage checks: projects/eda-leakage-profiling-showcase
  • Modern NLP classification, retrieval, and generation workflow: projects/modern-nlp-pipeline-showcase
  • Ranking model training fundamentals: projects/learning-to-rank-foundations-showcase
  • Ranking API serving and contracts: projects/ranking-api-productization-showcase
  • Time-aware demand forecasting foundations: projects/nyc-demand-forecasting-foundations-showcase
  • Demand API with metrics and tracing hooks: projects/demand-api-observability-showcase
  • Agent-guided research loops for tiny language models: projects/autoresearch
  • Agent routing, tools, guardrails, traces, eval rubrics, A2A/session/memory concepts, and SDK comparison: projects/agentic-course-assistant-showcase
  • Learned intervention control around a deterministic tutoring assistant: projects/adaptive-course-assistant-rl-showcase
  • Standalone capstone on where learning lives in an agent: orchestration-policy RL, offline RL and off-policy evaluation, an OpenAI Agents SDK bridge, RLHF/DPO/GRPO/RLVR, MARL, and an optional NumPy DQN/PPO lane: projects/learning-agents-showcase
  • Contextual bandits, MDPs, dynamic programming, tabular Q-learning and SARSA, REINFORCE policy gradients, optional DQN/PPO comparison, reward design, and deployment caution: projects/student-support-rl-showcase

First Run Pattern

cd projects/<project-name>
make help

Then run the recommended quickstart in that project's README.

Use your project's Makefile for its checks. The root commands below check the whole collection and require all project environments and generated artifacts. Harness commands are optional contributor tooling.

make check-contracts
make check
make verify
make docs-check
make harness-preflight
make harness-lint
  • make check-contracts regenerates missing or stale supervised artifacts and checks their structure and source/configuration and output hashes.
  • make check runs lint, type checks, tests, and contract verification across projects.
  • make verify checks generated artifacts across all projects and fails when required outputs are missing.
  • make docs-check runs a strict MkDocs Material build for docs consistency.
  • make harness-preflight and make harness-lint validate the repo-local public harness-lite bootstrap.

Docs Site

Run local docs with:

make docs-serve

Build static docs output with:

make docs-build

Topic Coverage Guide

For a direct mapping from course topics to projects, commands, and artifacts, use:

  • docs/aspect-coverage-matrix.md