REAL-LAB-GIST

REAL-Lab @ GIST

Real-world
Efficient
AI
Learning LAB

We bridge the gap between theoretical AI and real-world impact.

We bridge the gap between

theoretical AI and real-world impact.

Research Areas

Our work runs along three connected fronts: the mathematics that makes generative models work, the systems that carry them into the physical world, and the sciences they can accelerate.

Generative AI

We study why generative models work — the theory behind diffusion and flow matching, and the reinforcement learning that steers them toward what we actually want. The same foundation extends to language models and to agents that reason and act.

Theory of Diffusion & Flow MatchingRL for Generative ModelsLLMs & Agentic AIMultimodal Generation

Physical AI

Intelligence that has to hold up outside a benchmark. We build vision-language-action models, world models that predict how a scene will evolve, and the perception stack that grounds both in a real environment.

Vision-Language-Action (VLA) ModelsWorld ModelsComputer Vision & PerceptionGenerative Models for Robotics

AI for Science

Generative models turned toward discovery. We work on medical imaging and clinical decision support, and on molecular and protein design — domains where a good prior over structure is worth more than raw scale.

Medical ImagingHealthcare AIDrug Discovery & Protein DesignScientific Discovery