18-290 Signals and Systems

AI/ML-based statistically rigorous analysis of brain activities using EEG, supervised by Prof. Pulkit Grover (Spring 2024)

Enhanced Epilogue, an advanced acoustic dual-string barrier evaluation tool by incorporating heterogeneities in cement, resulting in a 3% improvement in homogenous accuracy and providing an uncertainty metric. Developed Physics-informed methods based on multi-task learning with domain generalization and ordinal regression to generate high resolution ultrasonic-like features from lower resolution sonic borehole data

Automated the control process of RF phase in a 15UD Pelletron accelerator facility for generating a stable pulsed beam in one of the first attempts to bring intelligence to particle accelerator operation in the country

Collaborated with a team of 4 interns for bringing user profiles from cloud to edge servers employing sequence modeling, autoencoders and multi task learning on user interaction data, creating low 16-dimensional universal user vectors. Adapted to the limited storage constraints of the edge servers and improved on latency, compression and updatability resulting in a patent, Adobe P10960-US (US Patent App. 17/849, 320)

SURGE Program