We stay ambitious, aspiring to introduce transformative changes and create an enduring legacy in the world.
We stay humble, recognizing that even tiny true progresses require unimaginable dedication, perseverance, and humility.
Building intelligent systems that autonomously advance scientific knowledge
We focus on AI methods development for computational biology and healthcare applications, especially recent Agentic AI development and its potential in scientific discovery. We believe one of the most important factors in data-driven biomedical applications is that the learned knowledge is not just specific to datasets or experiment protocols.
Our research spans from developing novel agentic AI architectures that enable large language models to function autonomously in complex data scenarios, to creating practical tools that deliver innovations to domain experts free of technical barriers. We are constantly talking to motivated students interested in these topics, either about admission or virtual research collaborations. Feel free to reach out.
Understanding the genetic basis of human complex traits
We study computational biology because every progress we make has the potential to free millions from suffering. We are devoted to developing methods that help understand the genetic basis of human complex traits. Our previous studies mostly focus on Alzheimer's disease and cancer, with ongoing work exploring connections between autism, neurodegenerative conditions, and other complex diseases through innovative computational genomics approaches.
Building trustworthy autonomous AI systems
We study the core ML topics such as agentic AI and its security issues, with a focus on building novel agentic AI architectures that enable LLMs to automate and function in complex data scenarios. We also study the issues of its security, protecting AI from being maliciously used. Our work includes developing transparent multi-agent systems and ensuring AI systems remain reliable, interpretable, and safe in real-world applications.
From this line of work we built TopCited.AI, a generative engine optimization (GEO) platform that helps brands get recommended and cited by AI-powered search engines and large language models such as ChatGPT, Gemini, and Grok.
Tackling the problems others say can't be solved
We pursue research directions that might be deemed overambitious or unconventional.
Heelo.dev is one of them: a new agentic AI architecture, built by a university lab to compete with the best that the largest AI companies offer. When we first released it in May 2026, it outperformed the strongest system of the time, GPT-5 with its agent harness, at about 10% of the cost. By October 2026 it was going head to head with Claude Fable 5 in Claude Code, arguably the most capable coding combination in the world, on coding benchmarks, at a lower cost.
The people, papers and posts behind this work.