Swapnil Sharma

I am an AI/ML engineer focused on building production LLM systems, multi-agent architectures, and evaluation infrastructure.

I am currently a Senior Research Engineer at Aleph, where I work on a semi-autonomous FP&A agent that helps finance teams reason over financial data, surface recommendations, and take actions with human approval workflows. My work spans agent capabilities, LLM-as-a-judge evaluations, and production reliability/performance improvements.

Before that, I was an AI Engineer / Senior AI Engineer at HippocraticAI, where I was part of the founding team and helped build the Polaris multi-agent system for safety-critical patient conversations. I worked on specialist agents, RAG/tool-calling, evaluation frameworks, and LLM fine-tuning, and helped scale the platform to 1M+ patient calls per week.

Earlier, I worked at Microsoft on code generation for JavaScript unit tests using OpenAI Codex, at Nference on biomedical search and NLP systems, and at Zeta (Directi) on high-throughput payments infrastructure.

I completed my MS in Computer Science at NYU, where I worked on code generation and bias evaluation research with publications on arXiv.

Outside of work I am building: lekh.space — a private encrypted writing vault; an MBTI-aware voice journaling app on LiveKit and Gemini; and Beacon, an emoji-only messenger.