I build probabilistic models that learn to generate solutions to hard combinatorial problems. Focused on GFlowNets, causal inference, and structured decision-making.
I'm a computer science master's student with a background in economics, working at the crossroads of probabilistic modeling, deep learning, and algorithm design.
My research centers on Generative Flow Networks (GFlowNets) — a new class of generative models that learn to sample diverse, high-quality solutions proportional to a reward. I apply these ideas to combinatorial optimization, causal structure learning, and Bayesian reasoning.
Theoretical foundations and practical applications of GFlowNets
Dynamic equilibrium models and Bayesian belief revision
Learning samplers for NP-hard problems via GFlowNets
Denoising-based generative processes and score matching
Open-source work at the frontier of probabilistic machine learning and combinatorial optimization.
Latent Unified Causal Inference through Dynamic Equilibrium — a GFlowNet-based framework for causal structure learning and Bayesian belief revision.
GFlowNet with critic for partition function estimation, applied to bilevel knapsack problems with Benders decomposition.
A GFlowNet that learns probabilistic solutions to the 0-1 Knapsack problem, enabling efficient global optimization via learned CDFs.
Exploration of diffusion models — implementation and experiments with denoising-based generative processes.
An unconventional path from economics to the frontier of machine learning research.
Specializing in machine learning, GFlowNets, and probabilistic modeling
Foundation in quantitative analysis, game theory, and optimization
Publishing code and frameworks on GitHub for the ML community
Interested in GFlowNets, causal inference, or combinatorial optimization? I'd love to hear from you.