My research interests center on efficiency across models, algorithms, and systems. I am specifically interested in bridging the gap between novel theoretical frameworks and low-level implementation details to achieve this.
Previously, I worked on efficient deep learning during two international research stints. At the University of Manchester, I developed a Multi-Agent RL communications framework that allows for backpropagation through discrete, unbounded communication channels (and currently working on extensions to image generation, world models and unsupervised learning). As a DAAD-Wise Fellow at TU Dresden, I focused on State-Space Models, developing a novel parametrization for Mamba to encode sparse geometric data from point-clouds and neurmorphic event streams.
My regular train from Langebrück to Dresden Hbf
Currently, I am a Performance and Research Engineer at e6data, working on database internals to optimize the core planning and execution engine.
For legal reasons this is a joke
I graduated from BITS Pilani with a B.E. in Computer Science and a minor in Data Science. During my undergrad, I explored a wide breadth of systems and ML problems: I designed cluster-aware file striping frameworks for parallel file systems at DaSH Lab and worked on pruning neural networks using Granger Causality at APPCAIR. I was also the Vice-President of SAiDL, the AI club on campus.
Suiting up for batchsnaps
Off the clock, you'll find me solving logic puzzles like Advent of Code or listening to Radiohead and BMTH. I've recently started cycling (survival record: 60km) and eagerly follow F1. I am was patiently waiting for Luka, Kyrie and the Mavs to take the NBA title.
Weekends