ML Infra

Advisor: Prof. Sayan Mitra

• Investigating indistinguishable sets computation for optimal SLAM in hybrid control system.

• Parallelized SMT solvers across GPUs to optimally compute indistinguishable sets for high-dimensional systems.

• Developed methodology to train and verify neural networks to compute indistinguishable sets for nonlinear systems, using SMT-verification and solver layers. Discovered major blindspots in car racing controllers.

Worked with Atom DA FE team, developing tools for CPU model performance evaluation which are frequently used by DA team currently.

Advisors: Prof. Talia Ringer, Hannah Leung

• Contributed to formal verification of Path ORAM.

• Designed memory access functions ensuring privacy-preserving RAM operations and proved correctness with Coq.