I build agentic systems that reason over real data.
DataVoyager
I led development of DataVoyager, a production-scale multi-agent system for data-driven scientific discovery, used by researchers globally, built in collaboration with the Allen Institute for AI.
NeuroDiscoveryBench
I helped build NeuroDiscoveryBench, a benchmark that evaluates how AI systems analyze neuroscience, genetics, and immunology datasets to generate scientific insights.
Alongside research systems, I’ve built LLM-driven insight platforms for high-volume production data and domain-specific retrieval systems for expert users, combining structured pipelines, RAG, and tool-augmented agents to generate decision-grade outputs.
DiscoveryBench (ICLR 2025)
Built all benchmark agents and the full experimental pipeline.
[Paper] [Github]
Data-driven Discovery with Large Generative Models (ICML 2024)
Contributed to system design, dataset construction, and experimentation.
[Paper]
Microsoft Autogen
Improved agents and the orchestrator by refining how they execute code, handle context and memory, reflect on failures, retry intelligently, and coordinate asynchronously across multi-agent workflows. See all contributions.
All-Hands-AI OpenHands
Integrated DiscoveryBench for scientific evaluation and improved inference and evaluation stability. See all contributions.
Website: https://abhijeetmeena.com
Contact via website form.


