a few things i've built across ai, ml, data, and systems.
an llm agent with an eval harness, gradient-boosted fuel-burn models, and an incremental data lakehouse, all over one set of flight data.
a live pipeline over public conflict data that runs about 70x faster than a full-scan baseline at 10m rows.
tenant isolation, rbac, rate limiting, and idempotent stripe webhooks, load-tested across concurrency levels.
a deterministic go exchange sim that measures how much trader latency quietly changes execution quality.
health diagnostics with real-time computer vision and neuro-data, from prototype to something running in clinical pilots.
a crisis platform that handled 33m+ hits in six weeks at zero downtime, with a chatbot that reconnected families.