Projects

AST-guided fine-tuning and retrieval-augmented generation for synthesizing Standard Verification Rule Format (SVRF) code from natural language, evaluated on a 741-example DRC benchmark and published at IEEE ICLAD 2025.

GPU-accelerated Inverse Lithography Technology (ILT) and model-based Optical Proximity Correction (OPC) optimization for sub-14nm semiconductor manufacturing nodes.

Fourier Neural Operator surrogates for semiconductor process simulation (etch, deposition, CMP), built at Siemens EDA to make TCAD process-window exploration interactive rather than overnight.

AI agentic flows and geometric engines to automate complex EDA physical verification and PDK validation.

Deep neural network pipelines and interactive dashboards to classify, filter, and clean Scanning Electron Microscope (SEM) data for chip metrology.

Match-winner classifier for men's professional tennis, built on 16,049 filtered matches and 12 engineered features, deployed as an interactive Dash application.

Two CS7641 studies benchmarking randomized search heuristics on discrete problems and neural network weights, and clustering with dimensionality reduction on stock and tennis datasets.

Automated multi-objective optimization and geometric programming routines for analog integrated circuit sizing.

High-throughput Density Functional Theory (DFT) simulations and equivariant GNN modeling for material adsorption energies.

Deep reinforcement learning models and non-stationary policy optimization engines for automated financial trading.

CS7643 Deep Learning team project combining pre-trained Transformer embeddings with convolutional feature extraction to classify news articles as reliable or fabricated.

Design study for an AI-first operating system: a polyglot microkernel with a Rust core and Python orchestration, specified around the scheduling problem that GPU-bound inference creates. Not yet built.