Research

Research Philosophy

I believe the next generation of AI will draw deeply from physics, taking inspiration from biological systems not just as metaphors but as computational principles. My research asks: How can we build systems that learn as efficiently as nature, reason about the physical world, and run on hardware that respects thermodynamic limits?

Focus Areas

Efficient LLM Inference

At Ludwig Computing, I work on probabilistic computing for efficient large language model inference, bringing together my background in stochastic neural networks, physics-inspired algorithms, and hardware-algorithm co-design.

Spatial Intelligence & Robotics

My previous work at Transio AI focused on spatial intelligence and world models: AI that understands and reasons about 3D space, spatial perception, embodied reasoning, and the foundations of Physical AI for autonomous systems and robotics.

Bio-Inspired Energy-Efficient Learning & Inference

Traditional deep learning is energy-hungry by design. I worked on alternatives to backpropagation and probabilistic, brain-inspired computing paradigms including binary stochastic neural networks, Boltzmann machines, and learning rules derived from statistical physics that can achieve competitive accuracy at a fraction of the energy cost.

Physics-Inspired AI & Optimization

Physics is not just a domain for AI to learn about. It can be a substrate for computation itself. My work on quantum-inspired optimization and Ising machine solvers demonstrates how physical phenomena can be harnessed for optimization and inference at scales beyond what conventional hardware supports.

Hardware-Algorithm Co-Design

From custom CUDA kernels for MCMC-based optimization to FPGA-based Boltzmann machine accelerators. I have hands-on experience bridging algorithms and silicon.

Experience