I am a Data Scientist at Zoox working on Perception Verification and Validation for autonomous driving.
Experience
Data Scientist, Perception V&V — Zoox (2026–Present)
- Driving statistical approaches to perception verification and validation for autonomous driving
AI Verification Engineer — Zoox (2022–2026)
- Designed and executed statistical approaches to provably safe autonomous driving
- Built tools, automated processes, and data products using Python, Databricks, and PySpark to support autonomous driving evaluation
- Designed data-collection strategies, fleet simulations, and test asset orchestration that enabled scaling and release of the Zoox service to the public
- Drove continuous improvements resulting in over 99% uptime
- Contributed to 9 patent applications (2 granted)
Graduate Research Assistant — McGill University (2016–2022)
- Trained, validated, and deployed models for dimensionality reduction, regression, and statistical inference on several hundred gigabytes of high-dimensional data
- Improved computational study design performance by over 50% using advanced statistical algorithms for design spaces robust to inactive dimensions
- Executed and managed 1,000,000+ simulations and 500+ CPU years of computing resources at national HPC facilities
- Developed and automated C++ and Python scientific-computing tools, streamlining workflows and reducing manual job orchestration
Education
- Ph.D. in Physics — McGill University (2022) Thesis: Quantification of the Quark-Gluon Plasma with Statistical Learning
- M.Sc. in Physics — McGill University (2018)
- B.Sc.(Hons.) in Physics, German Studies minor — The College of William & Mary (2016)
Skills
- Probability & Statistics: Bayesian inference, Uncertainty quantification, Experimental design, Sensitivity analysis, A/B Testing
- Languages: Python, Rust, SQL, Bash
- Data & ML: scikit-learn, Pandas, NumPy, SciPy, PySpark, Databricks, MCMC, SALib
- Software Engineering: Git, GitHub Actions, Bazel, Airflow, Flask, FastAPI, code review, software testing, CI/CD
- Open source: Author of maxpro — a Rust crate and Python package for Bayesian max-projection design of experiments
Selected publications
- M. Heffernan, C. Gale, S. Jeon, and J.-F. Paquet. “Early-times Yang-Mills dynamics and the characterization of strongly interacting matter with statistical learning.” arXiv preprint 2306.09619 (2023). arXiv
- M. Heffernan, C. Gale, S. Jeon, and J.-F. Paquet. “Bayesian quantification of the quark-gluon plasma with non-Gaussianinae priors.” Phys. Rev. C 109, 024912 (2024). arXiv
- For a full list, see my academic page or Google Scholar.
Patents
- 9 patent applications related to autonomous driving safety validation (2 granted)