Built an AI coaching copilot that embeds organizational culture directly into employee workflows. Started with simple prompt flows and evolved the architecture into modular agent-based systems that deliver personalized behavioral coaching during real-time work: emails, Slack, meetings; aligned with each company's values. Optimized latency, infrastructure costs and accuracy through prompt optimization and smart caching. Designed evaluation pipelines and synthetic data generation that accelerated iteration cycles. Shipped to enterprise clients serving 10k+ employees each, making HR coaching actually work at scale.
25
204
Built production infrastructure for an AI assistant that captures screen context and meeting data to create a persistent memory layer for knowledge workers. Developed content deduplication pipelines, optimized LLM parsing systems, and designed modular agent architectures that power contextual chat, meeting transcription, and proactive task suggestions. Shipped systems serving real users who rely on LittleBird as their "second brain.
2
68
Built AI systems that turn hundred-page medical records into structured case intelligence for personal injury law firms. Architected context-aware agentic workflows that maintain state across complex documents, solving the core problem of extracting precise categorizations and structured data from messy litigation files. Developed custom benchmarking frameworks and entity resolution pipelines validated against production datasets, achieving 100% recall. Delivered infrastructure that helps leading PI firms save 10+ hours per case and accelerate settlements.
1
63
Led the build of a GenAI tool that makes financial analysts faster, partnering with Lynott Partners hedge fund. Designed the whole stack AI pipelines, backend, QA; from research papers to production code. Scaled the team from 2 to 7 as Interim CTO, took the company to $5M valuation. Multi-agent systems, advanced RAG, real financial insights. Made finance less painful.
1
89