Projects that I have worked on over the years.
Simulated quantum circuits with stochastic Pauli error channels using cuQuantum tensor-network statevector backends, developing statistical sampling algorithms to isolate gate errors under NISQ regimes.
Ultra-low-latency video generation pipeline utilizing LTX-2.5 Diffusion Transformer (DiT) on NVIDIA RTX 5080 GPU, featuring custom NVFP4 quantized GEMM kernels and asynchronous pinned-memory weight offloading.
Designed and trained conditional diffusion generative models to sample continuous molecular dynamics and biomolecular trajectories on Slurm HPC, bypassing expensive QM/MM simulations (published in ACS JCTC).
Architected an extensible AI development environment integrating Model Context Protocol (MCP) to orchestrate multi-step tool-calling workflows, backed by a scalable agent runtime on AWS EC2 and optimized edge SLM inference.
training LLaMA3.2 to predict idea density scores of text
