I build AI products and developer tools with Python and TypeScript, focusing on LLM evaluation, reliable workflows, and useful interfaces.
Based around Waterloo / Toronto · Interested in applied AI / LLM engineering roles · Website & background
| Project | What I built | Evidence |
|---|---|---|
| Zeta Reason | LLM evaluation tooling with FastAPI and React: dataset-driven comparisons, experiment history, and per-task output inspection. | Code & setup · Evaluation caveats |
| PagePulse | Exact-URL monitoring with deterministic change detection, optional LLM summaries, email alerts, and primary/shadow evaluation methods. | Architecture & engineering decisions · Private source |
| H200 Kernel Lab | GPU benchmarking, integration, and validation of BF16 projection kernels. AI-assisted work with explicit upstream attribution. | Saved official result: 1.52067× baseline; an adaptation of public kernels, not a claim of novel kernel authorship |
| Docs Drift Check | A deterministic Git-diff scanner packaged as a Codex skill/plugin, with an agent review workflow for likely documentation updates. | Demo · Validation scenarios |
| GCN reproducibility study | PyTorch GCN vs. MLP on Cora, with multi-seed evaluation and a label-sparsity study. | Stored results: 80.36% vs. 55.40% mean test accuracy over five seeds |
- Evaluations that expose failure cases and distinguish measured results from proxies.
- AI features built around deterministic checks and inspectable evidence.
- Product workflows spanning APIs, workers, data storage, and frontend interfaces.
Applied ML / Computational Mathematics at University of Waterloo; product and analytics work at Volaris / Aislelabs.
Python · TypeScript · PyTorch · FastAPI · React / Next.js · PostgreSQL · Supabase · GitHub Actions