Episode 63: Building AI Agents That Learn Insights from Elman Mansimov, Research Scientist at DeepMind

About the Episode

In this episode of the AI Agents Podcast, we dive deep into the evolving world of AI agents with Elman Mansimov, a research scientist at DeepMind and founder of Sourcely. Elman shares how his background in generative AI research—from early text-to-image models to pioneering work in machine learning at the University of Toronto and NYU—has shaped his vision for building intelligent agents that can truly learn, adapt, and reason. He discusses the challenges and breakthroughs involved in developing Sourcely, an AI-powered academic tool designed for deep search and source verification across diverse fields from fitness to academia. Explore how Elman sees the future of autonomous agents transforming everything from academic research to software development, and why context length, reasoning capabilities, and user personalization are key to building smarter AI tools. Whether you're an AI enthusiast or a product builder, this thoughtful conversation reveals what it really takes to build agents that learn and deliver value in meaningful, nuanced ways.