A portfolio from the Duke University AI Product Management program. Six workstreams — MVP design, build-vs-buy economics, evaluation frameworks, bias & fairness, MLOps data labeling, and strategic roadmapping — paired with runnable Jupyter notebooks so the methodology is code, not just claims.

Applied at production scale in explanova.ai — a K–12 S.T.E.M. homework copilot I solo-built using these exact frameworks.

For mathematical detail on the evaluation criteria and data architecture decisions, see TECHNICAL.md.

Three notebooks in /notebooks execute the frameworks on public/synthetic datasets:

cd notebooks
pip install -r requirements.txt
jupyter notebook

All three run in under a minute on a laptop.

The /web folder is a Vite + React 18 + TypeScript + Tailwind site that renders this portfolio at stephengardnerd.github.io/AI_Product_Management. GitHub Actions deploys automatically on every push to main that touches web/.

The end-to-end flow this portfolio was built against — discovery to delivery, with the workstreams each section of the repo exercises:

gantt
    title AI Product Lifecycle
    dateFormat  YYYY-MM-DD
    axisFormat  Week %V
    section Discovery
    Hypothesis generation       :done,  a1, 2025-01-06, 7d
    Build vs. buy analysis      :done,  a2, after a1, 5d
    section Architecture
    Data provenance & ethics    :done,  a3, after a2, 7d
    Human-in-the-loop scaling   :done,  a4, after a2, 10d
    section Delivery
    Evaluation & metric mapping :done,  a5, after a4, 5d
    Roadmap & MVP delivery      :done,  a6, after a5, 7d

_{Author: Stephen D. Gardner · @stephengardnerd · explanova.ai}

_{MIT Licensed}