🏢 Research Assistant at UNC Chapel Hill
🎓 Master's in Computer Science (Aug 2024 - May 2026)
🌱 Working on formal verification for RISC-V architectures

Swift RISC-V INT8 Accelerator

• Designed a high-performance RISC-V coprocessor in SystemVerilog, integrated with the CVA6 core
• Utilized the CVXIF interface to offload INT8 operations, achieving significant throughput increase
• Optimized neural network inference latency through hardware-software co-design

AI Predictability & Assessment Toolkit

• Designed a comprehensive ecosystem encompassing React, Node.js, and Browser Extension
• Employed a hybrid measurement engine using LLM-based proxy classification, Shannon entropy, and KL divergence
• Created a recommendation engine connecting complex predictability measures to practical design advice

Clinical Trial Matching System

• Developed an end-to-end recommendation system using Python and LLMs, matching cancer patients with clinical trials
• Built a RAG (Retrieval-Augmented Generation) pipeline using Vector Databases and FastAPI
• Implemented medical Chain-of-Thought (CoT) reasoning for explainable eligibility assessments

Distributed Consensus Testing

• Checked the Raft consensus protocol for security issues using TLA+ formal modeling and KLEE symbolic execution
• Developed a C-based symbolic test harness to replicate message drops and network partitions
• Utilized a dockerized environment to automate state-space exploration and validate protocol invariants

• Resolved high-concurrency race conditions by implementing asynchronous locking for thread-safe usage tracking
• Currently collaborating to architect a universal structured citation system supporting OpenAI, Anthropic, Google/Gemini, and Perplexity providers

• Developed special checks for NVIDIA Blackwell GPUs (SM100+) to fix INT8 quantization issues
• Added automatic backup options for FP8 and enhanced kernel loading mechanisms

• Fixed design issues in the CVA6 core by correcting RV64 interrupt cause recording
• Stopped unnecessary latch creation in the logic, improving hardware reliability and stability

• Developed a diagnostic suite to identify resource leaks in the Streamable HTTP transport
• Highlighted risks of connection pool exhaustion during failed SSE streaming events