Primary engineer on an AI dream-journal app and its Python backend, authoring about half of all commits across both codebases (3,300+ mobile, 700+ backend).
- Architected a multi-agent AI chat on LangGraph that routes each message to specialist agents and streams replies in real time over SSE, with locking and rollback so partial failures never leave inconsistent data.
- Built the retrieval & memory layer: a Pinecone RAG pipeline with token-budgeted context, long-term memory ranked by time-based decay, running conversation summaries, and content moderation.
- Created an LLM evaluation harness — classifier precision/recall and LLM-as-judge groundedness checks — to compare prompts and models before release.
- Designed Supabase/Postgres schema & migrations (RLS, triggers, row locking) and Redis/RQ background jobs for indexing and summarization.
- Shipped mobile features end to end in React Native/Expo + Redux Toolkit, plus native iOS (Swift) and Android (Kotlin) home-screen widgets.
- Integrated RevenueCat, Apple/Google sign-in, push notifications, PostHog and Sentry; wrote 200+ test files and 22 Maestro E2E flows.
LangGraphFastAPIReact NativePineconeSupabaseRedis