About

I was jack of all, master of none. Yet, I dared do a PhD. For what is PhD, but philosophy. And philosophy i did. Turned out jack is still trying to master becoming jack.

I am currently pursuing PhD in computer science at University of Central Florida. I am being advised by Dr. Yogesh Rawat and Dr. Mubarak Shah. Prior to my PhD, I worked as Senior Member of Technical staff at Athena Health. Before that, I worked at Samsung Research under Dr. Manish Sharma. In an earlier life, I spent time at IIT Roorkee, Parimal Lab, under Dr. Partha Pratim Roy. During my undergrad at NSIT, I was fortunate enough to have worked and learnt from Dr. Harish Parthasarathy. I will forever cherish the time spent discussing quantum computing, and various mathematical problems with Dr. K.R. Parthasarathy. Last, but not the least, i had the great privilege of learning flute from my Guru Rajendra Prasanna.

Research Interests

For a long time, i have been interested in building compute-efficient AI models that adapt under “distribution shifts”.

From an industry standpoint: I am interested in deploying AI models “on the edge”, specifically techniques like weight pruning, quantization etc. I am also searching for a way to transfer knowledge in a zero-shot way between architectures of different structures.

From a research standpoint: Now, i am just trying to figure out how the brain “might” work. I read neuroscience, and then build computational models encoding those constraints. My interest lies on building biological plausible learning algorithms, simulating cognitive thought processes, inventing new kinds of “associative” memories. More specfically, i am interested in lifelong learning, i.e. how to build machines which learn even during inference and do not rely on pre-defined SSL tasks.

Some questions which excite me: 1) what is next after backpropagation 2) How to encode part-whole hierarchies in neural nets 3) How to solve the “binding problem” in neural nets 4) How to build machines which can solve optical illusions like Necker’s cube, Kaninza Triangle, and Spinning Dancing Lady. 5) Build quantum superposition neural nets 6) Mechanisms of sleep in brain, and its role in memory storage 7) Fast-weights to store temporary neural activities 8) How to solve hard problems of consciousness. 9) How to invent memories which don;t generate neural activity at all, or directly compress these activations as soon as they are generated. These problems taste good to me.

The challenge then is which one to solve first. Now that i hold these problems in mind, i keep looking for cool tools which will enable solving them. And as soon as something appears, i focus on solving it.