Open Switch
LLM gateway with circuit breakers, half-open probes, hedged requests, cost-aware routing, per-key budgets, and request traces.
Work
Production LLM systems, Rust infrastructure, applied ML, and the full stack products around them.
LLM gateway with circuit breakers, half-open probes, hedged requests, cost-aware routing, per-key budgets, and request traces.
Typo tolerant hybrid search with faceted filters that runs fully in the browser.
LLM evaluation runner with its own YAML check format, flakiness detection, and a pull request gate that writes a Markdown report.
FX exposure and hedging engine with a quantile forecast that is honest about what it cannot predict.
Multilingual AI agent that answers student enquiries, fills a CRM, and hands off decisions to a counselor.
Schema-first code generator where deterministic structure keeps the LLM focused on application logic.
React library for turning chat buttons into explicit decisions.
Tested the Platonic Representation Hypothesis at toy scale. Reproduced a reference alignment method to 96.37% against a published 96.01%, then recovered held-out structure between independent encoders after a centering ablation.
Synthetic stress test of a venue savings score, with an audit of confounding, observability, and tuning claims.
React and Django store with Stripe PaymentIntents, order tracking, and email notifications driven by Django signals.
Interactive 3D building configurator in the browser.