Professional Experiences

Highlights

• Co-authoring research paper on automated multi-agent security vulnerability remediation, contributing pipeline architecture and end-to-end evaluation framework across 120 Java CVE benchmarks.
• Designed and evaluated a 4-stage agentic security pipeline (Finder → Exploiter → Patcher → Verifier) integrating IRIS/CodeQL static analysis with LLM-driven exploit and patch generation, benchmarked end-to-end on 120 Java CVE projects.
• Reduced pipeline iteration latency by 72% and cut redundant static-analysis recomputation by 85% by implementing a state-based SARIF caching layer for the Finder stage of a LangGraph-based agentic pipeline.
• Increased pipeline coverage from 73% to 100% across a 120-project Java CVE benchmark by introducing graceful fallback handling for missing cache artifacts, eliminating startup failures on previously-unseen projects.
• Reduced redundant analysis on retry paths by 40% by decoupling source-tree extraction from cache invalidation in the Finder dispatch logic, collapsing 4 incoherent flag-state combinations into 2 well-defined modes.
• Cut per-experiment setup time by 85% by refactoring CLI flag semantics, eliminating manual code edits previously required to toggle between cached and fresh pipeline modes.

Projects

Projects

Highlights

• Researched attention mechanisms and transformer architectures, focusing on how large-scale matrix multiplications dominate training and inference of LLMs.
• Authored a 'read-along guide' on GPU-accelerated matrix multiplication kernel optimizations in CUDA, outlining concepts based on GPU multi-threading and execution models (thread blocks, warps, SM scheduling).
• Leveraged CUDA programming to design kernels in Linux environments, achieving 93.7% of cuBLAS using NVIDIA Nsight Compute (NCU) for analysis. Inspected PTX assembly output to further unravel optimizations, understanding IR passes done by the compiler itself.

Highlights

• Taught programming and robotics fundamentals using Python, Arduino, and microcontroller labs, enabling students to successfully program motors and sensors and complete functional robot prototypes.
• Led student robotics projects from concept design through testing, guiding teams in mechanical iteration and debugging with programmable hardware, resulting in completed robots that met project goals and were showcased in a school exhibition.

Highlights

• Led weekly peer-led sessions for a dedicated group of students, fostering collaborative learning and long-term academic growth.
• Explained key concepts through interactive discussions and guided problem-solving sessions.
• Reinforced personal mastery of calculus through teaching, feedback, and real-time troubleshooting.

Highlights

• Organized and led events to welcome and engage new students.
• Collaborated with other Hive Leaders to plan and execute group activities.
• Supported and mentored HIVE volunteers aspiring to become Hive Leaders.
• Led orientation sessions, ensuring new students had a smooth transition.

Highlights

• Front Desk Reception: Managed check-ins, answered inquiries, and kept daily operations running smoothly.
• Customer Service: Built relationships with clients, addressed concerns, and made sure every member felt welcomed.
• Fitness Class Assistant: Supported instructors with class setup and equipment while keeping participants motivated and sessions running safely.
• Intramural Soccer League Assistant: Helped organize and run a soccer league — handling scheduling, officiating, and ensuring games ran smoothly.