Notebook

Qualcomm

Built and deployed a production application for 200+ users to bring automotive product, configuration, and project information into one place.

• Developed agentic search across projects so users could find comparable products, compare technical specifications, and retrieve relevant information across customer engagements.
• Consolidated information from multiple sources to support product evaluation and technical requirements alignment.

University of Washington Math AI Lab

Worked on mathematical theorem retrieval across natural language, LaTeX, and Lean.

• Developed a fine-tuned embedder with guidance from Vasily Ilin and Henry Kvinge.
• Presented the work at the ICML 2026 AI for Math Workshop; accepted at the NeurIPS 2026 Math AI Workshop.

Advanced Robotics at the University of Washington

Optimized a YOLO-based detection and classification pipeline on NVIDIA Jetson, reducing end-to-end latency from roughly 30 ms to 6–7 ms.

• Used TensorRT, CUDA, quantization, and GPU-accelerated preprocessing and postprocessing.
• Improved model execution and CPU–GPU memory transfers. ARUW placed third at the 2026 ARC Robotics Competition.

University of Washington

Computer Engineering; previous attendance.

Poway Unified School District, KasmV2

Initiated and engineered a multi-server AWS platform delivering browser-based desktops and applications across five high schools.

• Built Docker and Nginx infrastructure with load balancing, AWS Auto Scaling, and Ansible/Terraform provisioning.
• Automated deployment and centralized access to containerized computing environments.

TrinamiX

Trained machine learning models to identify materials from laser speckle patterns using fractal dimension calculations.

• Built a custom dataset pipeline using polarized lenses and high-resolution imaging to capture speckle samples.

LLM Matching Research Project

Explored LLM-based conversation simulation and matching under the mentorship of Stanford PhD student Naman Gupta.

• Optimized Gale–Shapley-related LLM calls using lazy pool optimization and clustering, with a reported 70% reduction in compute time.

Cambridge Centre for International Research

Worked on Hindi–English code-switched translation using machine learning and NLP.

ColorTokens

Built automation for Linux security services and used Python to automate log ingestion and anomaly detection.

• Supported vulnerability detection and remediation, reducing the need for manual review in server-network workflows.

CyberPatriot

Captained a two-time national finalist and national champion team in Linux, Windows, and network security.

• Worked on vulnerability remediation, patching, service hardening, and live defensive exercises.

FIRST Tech Challenge, Team 15171 Scorpio

Led programming and team strategy, developing vision-based autonomous behavior and navigation for competition robots.

• Worked with OpenCV, neural networks, odometry, PID control, and finite state machines.
• Dean’s List Semifinalist in 2022 and 2023.

Tech Space

Introduced peers to emerging technologies through workshops, demonstrations, and collaborations with engineers.

Aerospace & Rocketry Club

Led projects involving drones, model rockets, and STEM outreach.