SciML Consultancy

Applying Foundational Artificial Intelligence Research
for Practical Scientific Challenges.

I seek to engage with organizations seeking independent, research-informed advice on complex scientific problems which can leverage ML. Typical cases where my efforts would be best suited are - if there is a need to teach modern ML models (transformers, LLMs, neural operators and PINNs) and/or to design them specifically for solving for the time-evolution of a high-dimensional system.

Dynamical systems that we currently think about are generally from the realm of (multi-)physics: those involving a flowing fluid - like ocean waves or blood flow in the arteries, time evolution of lattice models of magnetism and chemical reactions - to name a few of our active application areas. As a group specializing in the foundations of AI, we leverage a key insight that all such systems can be dealt with by fairly universal transformer based ML systems, and hence experiences and learnings from one transfer to the others.

Have a look at this demonstration of how we can design transformer to predict synthetic rough waves! This above model beats a lot of other published models on this particularly hard wave task! Contact us if you want insights on how to scale this up to your usecase!