Jason Maskell

Leading AI engineering at an early-stage startup building production LLM systems — agents that generate, decide, and respond in real time, running live across web and mobile. - Replaced the legacy production backend solo: ported the prior engine to a TypeScript monorepo, verified to parity with parity suites and mutation testing, and cut it over atomically with zero customer-facing regression. - Architected the platform on a contract-first design — schemas as the single source of truth for types and runtime validation at every boundary, no codegen, no drift. - Built the supporting services end-to-end: Go gRPC multi-provider LLM proxy, Go WebSocket STT proxy, a deploy control plane with atomic content releases, and a CI-integrated LLM evaluation harness with automated playtesting and LLM-judge scoring. - Designed the core LLM-runtime systems: staged generation pipelines with prompt caching and safe parallelism (~40% measured latency reductions) and streaming tool-use.

Working on an agentic-AI based game development startup called Visionary. Also doing codev work in Unreal Engine/various server-side tech stacks.