01 / RESEARCH
Better questions.
Deeper foundations.
Our work moves between mathematical structures and the systems they make possible.
The mathematics of learning
What changes when we treat learning as a mathematical object? We connect optimization, probability and geometry to study distillation, diffusion models and the structure of neural learning.
Theory of deep learning · Geometry · Optimal transport
Intelligence in the physical world
Models must eventually meet the world. Our work spans visual world models, robotic co-design, locomotion, manipulation and the geometry of control.
Computer vision · Robotics · Control
Learning systems, under scrutiny
A model is more than its benchmark score. We study evaluation, grounded tools and the trade-offs between model scale, memory and quantization.
LLM evaluation · AI infrastructure · Efficient inference
Mathematics into decisions
We bring structured models to scientific and financial questions: physics-informed learning, stochastic control, optimal stopping and market dynamics.
Scientific ML · PINNs · Financial AI
From an idea
to an experiment.
Explore code and research projects from the team.
SSVI Surface Calibration via Schrödinger Bridge
Gordei Verbii · Georgii Kilinkarov
RESEARCH CODE ↗Reinforcement Learning for Optimal Stopping with Financial Applications
Gordei Verbii · Georgii Kilinkarov
RESEARCH CODE ↗Scale vs. Quantization: Mapping the Memory Wall on Consumer Hardware
Nauan Darkhan · Gordei Verbii
RESEARCH CODE ↗Actuarial-Informed Neural Networks
Gordei Verbii · Ilya Starikov
RESEARCH CODE ↗OpenMines
Ilya Starikov