From Generative Text to Generative Discovery: Talk by Dr. Vinayak Agarwal 

On November 8, in our IITAGNE recent event in the Generative AI technical talk series, Dr. Vinayak Agarwal, Lead AI Researcher at BioState AI and founder of KDense, delivered an insightful talk on the future of AI-driven scientific discovery. A former mechanical engineer from IIT Bombay and PhD graduate from MIT, Dr. Agarwal shared his journey from traditional engineering to cutting-edge AI research, highlighting how agentic AI scientists are redefining the frontiers of science.

He began by tracing the evolution of generative AI, from text generation tools like ChatGPT to agentic systems capable of autonomous reasoning and scientific exploration. While early generative models unlocked creativity and productivity, Dr. Agarwal noted their limitations: hallucinations, lack of trustworthiness, and shallow reasoning. These challenges, he explained, have inspired the emergence of agentic AI, systems that combine reasoning, memory, and action to perform end-to-end tasks.

Dr. Agarwal discussed Model Context Protocols (MCPs) as a key innovation enabling AI models to process sensitive data securely while preserving user privacy, a crucial step for deploying AI in scientific and industrial contexts. MCPs, now widely adopted across platforms, allow data to stay local while AI models interact through structured, privacy-preserving channels.

A highlight of the talk was KDense (https://k-dense.ai/), an Agentic AI Scientist platform that automates the scientific discovery workflow. By mimicking how human researchers conduct literature reviews, design experiments, analyze data, and write papers, KDense dramatically shortens research cycles from months to hours. Dr. Agarwal shared how this system has already accelerated biological aging research in collaboration with Dr. David Sinclair, demonstrating AI’s transformative potential in the sciences.

The session closed with an engaging Q&A on the future of AI-enabled “vibe research” - the fusion of human intuition and AI reasoning. The discussion left the audience inspired about a future where AI is not just a tool for text generation but a true partner in discovery.

|