Krishna Yadav

I’m an AI Researcher and Founding Engineer with 4+ years of experience building multi-agent systems, RAG pipelines, and federated learning frameworks. My work has been published in AAAI, INFOCOM, and ICDM, with 500+ citations, and I’m motivated by turning research into production impact.

In industry roles, I led enterprise conversational AI efforts for the Polkadot ecosystem, designing RAG with semantic search, persistent memory, safety guardrails, and text-to-SQL for on-chain natural language queries. I’ve driven measurable performance gains—2× faster responses, 60% engagement lift across multi-platform personas, and improved proposal success rates by 40% through an autonomous multi-agent delegation system.

I also build end-to-end systems from data pipelines to model optimization, including GPU-accelerated constrained generation, fine-tuning and retrieval architectures, and robust telemetry-driven engineering. On the research side, I founded and lead SAI Lab and develop frameworks like FedGRAA, vertical federated learning, and federated machine unlearning—always with a focus on reliability, efficiency, and real-world security challenges.