🔬 From Generative Text to Generative Discovery: K-Dense and the Rise of Agentic AI Scientists
Join us for an insightful technical deep-dive on the next frontier of Artificial Intelligence: the transition from simple large language models to Agentic AI scientists that can autonomously conduct research and discovery.
Vinayak Agarwal, a leader in AI research, will demonstrate how his team is moving AI beyond drafting emails and summarizing documents into forming scientific hypotheses, planning complex multi-step experiments, and validating discoveries. This is a unique opportunity to learn the core design principles and practical application patterns for building AI that truly thinks and does science.
You'll discover how the K-Dense framework organizes AI discovery around structured reasoning, tool use, and self-critique—leading to immense gains in speed, quality, and reproducibility across scientific domains.
Speaker
Vinayak Agarwal is the Lead AI Researcher at Biostate AI, where he develops cutting-edge systems that enable scientists to use AI for autonomous research and discovery. He has spent years building state-of-the-art agentic AI systems, advancing the capability of machines to reason, plan, and explore scientific domains independently.
- Background: Vinayak earned his undergraduate and master’s degrees in Mechanical Engineering from IIT Bombay in 2018 and completed his PhD at MIT, focusing on auditory perception and cognition-inspired engineering. He has also contributed to large-scale AI initiatives during his internship at Meta.
- Beyond Work: He is a semi-professional Mohan Veena player, combining his interests in music, cognition, and intelligent systems.
Key Takeaways
Attendees will learn:
- The progression from simple LLMs to robust Agentic AI.
- The architecture of K-Dense, a cognition-inspired framework for AI scientists.
- How K-Dense systems generate high-value hypotheses, prioritize mechanisms, and produce research writing that stands up to expert review (including a collaboration with David Sinclair).
- Practical criteria for evaluating AI discovery: novelty, robustness, reproducibility, and experimental value.
Date: November 1st, 2025
Time: 12:00 to 1:00 PM EST
Location: Zoom. Meeting link will be sent to attendes.
