Working in the core query engine team. Making algorithmic improvements for better memory efficiency and CPU utilization to enable faster SQL query execution.
Organized a symposium, conducted courses, undertook research projects, hosted paper-reading sessions and hackathons.
State-space models for neuromorphic event streams and point-cloud data. Work accepted at NeVI @ ICCV'25.
Detecting user habits using CSI measurements from WiFi devices. Runner-up for best paper in WiP track at PERCOM'25.
Designing a cluster-aware file-level adaptive striping framework for parallel file-systems (BeeGFS). Published at IEEE CLUSTER'23.
Implemented a visual servoing algorithm using a custom-trained YOLO as the object detector, integrated PID-based GPS navigation with the rover's tech stack. 2nd Place in Autonomous Task at IRC'24. Project website.
Explored causal neural network pruning strategies that preserve feature importances, compared against L0, L1, Rank-based and Learning-Compression Algorithms. Code.
Used UNets and LSTMs for rainfall prediction using satellite data. Experimented with genetic algorithms to compare model performance against gradient-descent based optimisers. Work accepted as a poster at IGARSS'24.