I develop core systems to help doctors and nurses fill out complicated medical forms using patient data.* AI Technologies: LLMs (OpenAI/Anthropic), LangChain, Llama Index, Qdrant, OpenCV* Scalable Systems: Engineered production system using Temporal.io to ingest thousands of patient records concurrently, using LLMs to answer hundreds of detailed medical questions per patient.* Accuracy Improvement: Designed a system to measure and improve the accuracy of AI-driven question answering, through reinforcement learning from human feedback (RLHF) with healthcare professionalsMy work has significantly increased our system's throughput and scalability, transforming our processing capabilities from dozens -> thousands of patients, and transitioning our product from a proof of concept to a fully operational product with paying customers.

• Owned development of core features for Myxt's music collaboration platform on web and mobile (Flutter) including Notifications, Social Sharing, and AI powered song transcription.• Built user analytics pipeline, and used these insights to drive >2X weekly app store downloads• Ran experiments with coupons to drive user growth

• Reduced prevalence of harmful search results for ~2B MAU on Instagram Search, by developing safety and integrity metrics used as features in search ranking systems. • Led measurement for a global pre-launch experiment of the new Shop tab in Instagram app, and coordinated feature launches across multiple product teams to drive significant increases in weekly engage visitors.• Built early metrics for ML research teams to measure effects of Differential Privacy and Federated Learning, helping to improve performance of privacy compliant models.

• Built features for web (Angular, Ruby) and mobile apps for smart cooking devices. Worked with designers to optimize onboarding flow in complex wifi and bluetooth pairing UIs.• Responsible for data pipelines using AWS to store and analyze device logs• Built internal monitoring systems for hardware QC using log data (Elasticsearch)• Drove content decisions with customer usage data

• built frontend (React) and backend (Java) features for SMART's realtime classroom collaboration platform• integrated OAuth2 into SMART's automated textbook ingestion system

• used automation and computer vision to study signaling in stem cells.• developed methods to test ~100k compounds on human stem cells in order to discover novel therapeutics.• built javascript apps to visualize large biological datasets.

• designed and built CRISPR molecules to visualize stem cells.• collaborated with a partner lab to use these tools to identify affected cells in a study of Muscular Dystrophy