Academic & Research

Research interests, academic projects, and ongoing educational journey.

Education

My formal and informal educational journey across multiple institutions and disciplines.

Ongoing

Completed

Other Learning

Current Research Interests

My research interests sit at the intersection of artificial intelligence, neuroscience, and complex systems. Most of it just starts at curiosity and learning / understanding more about an area, and once in a while something ends up going deep enough to uncover something new that others might have overlooked, not looked hard enough for, or with the advent of super smart AI became tractable enough to attack from multiple angles fast enough to discover something useful.

Geometric Singular Learning

I am fascinated about the action, learning and understanding loop and one of several thought experiments, theory crafting and computational experiments about representational invariance led me to Singularity Learning Theory. It had the closest analogues to some of my ideas. At the same time I had also been looking into geometric deep learning approaches to machine learning. This confluence and refusing to stop going deeper at every roadblock led me to what I would call a pretty interesting bridge between Information Geometry and Singularity Learning Theory. Thanks to this rabbit hole I also ended up building a fully distributed training and experimentation system connected to hand assembled server workstation with 4xGPUs that have been increasing my homes temperature and power bill beyond what is safe to justify. Several dozens of 100M - ~1.4B parameter language and models, and hundreds of fine tunes of language and vision models form 10M to 32B parameters are behind... a lot of unreleased and in-progress work. I must say though, it has been intensely fun and enlightening.

The very foundation and first cut of a tiny part of all that work has made it ArXiv - Dead Directions: Geometric Singular Learning

World Models

My interest in world models primarily started with an interest in latent representations. World Models have been (were?) all the rage in early 2026. I've spent quite a bit of time building small world models bit by bit and a curriculum to teach myself the various building blocks and components that go into building world models. These have mostly remained at toy or symbolic levels, since video based world models are insanely compute intensive.

A very good survey of World Models is written by Pim De Witte and Packy McCormick if you want to get a lay of the land.

Neuromorphic Computing

Fairly interested and spent quite some time learning about Neuromorphic Computing. Prof. Chetan Thakur (one of the few academics in India interested in the area) at IISC covered a short module on campus when we were visiting at the end of the VLSI Design Program. It's fascinating but with real production challenges, waiting for the right projects where can add some neuromorphic computing jazz for efficiency wins.

Computational Neuroscience / NeuroAI

Fascinating mixture of understanding how our brain and computational systems come together. Completed 2 programs via Neuromatch in the summer of 2024 and 2025. Our NeuroAI project getting C-elegans to learn to switch gait and swim in different environments using a biologically inspired neural network - with barely a dozen neurons, was selected for the Neuromatch Impact Scholars program, where we took it a micro-publication level. With work in progress for a full publication submission.

Apart from various other new concepts I was glad to have found Peter Gärdenfors - Conceptual Spaces: The Geometry of Thought which I found to be an interesting approach of conceptual representations.

Hardware-Software Co-Design for AI

Another area of interest fed fuel via pain and suffering of dealing with imports of sensors, wearable hardware and programs at IISC (VLSI Design) and IIIT Hyderabad (Hardware in AI)

Couple of Projects

Current Affiliations

Publications & Preprints

Citation & Contact

For academic collaborations, paper discussions, or citation requests email with [RESEARCH] in the subject line: