Unified RAG pipeline for investment intelligence
Modular claim-first retrieval + grounded synthesis powering chat, memo generation, and research pages.
portfolio
Few-shot medical research classifier (200k records)
Automated classification with 93% human-machine agreement and a precision-first abstention strategy.
Real-time market monitoring pipelines
Operational health monitoring to reduce incident detection time and improve reliability at infrastructure scale.
takeaways
Why Technology Alone Does Not Solve Business Problems
Published:
Technology is an enabler — value comes from aligning AI systems to users, decisions, and workflows (and measuring failure).
talks
Concordia Contest NeurIPS 2024 — LLM-based Multi-Agent Cooperation
Published:
Research project on LLM-based agents in mixed-motive social dilemma environments, focused on cooperation, generalization, and experimental evaluation using Concordia.
AI Otherwise Workshop — Trustworthy Personal AI Agents
Published:
Workshop activity on participatory and speculative approaches to trustworthy personal AI agents, focused on human preferences, co-creation, and deployment infrastructure.
Citizen-Centric Artificial Intelligence Systems (CCAIS) Workshop 2025 – Multi-agent Negotiation Lightning Talk
Published:
Lightning talk on how LLM-based agents can support cooperation, conflict resolution, and collective outcomes in multi-agent negotiation settings using Concordia.