UC San Diego · Computer Science B.S. 2028
Undergraduate student (Chancellor's Honors, 4.0 GPA) specializing in computer vision, machine learning, and full-stack development with hands-on experience architecting AI/ML pipelines, autonomous robotics systems, and production software across research, technical leadership, and award-winning hackathons.
Projects
V2X Perception
CISL · UC Riverside
Developed a Cooperative Perception (V2X) framework that resolves sensor blind spots, such as pedestrians occluded by parked vehicles. Implemented MMDetection3D on a Roadside Unit (RSU) to perform real-time 3D object detection on live traffic streams, followed by a spatial perspective-transformation pipeline that reprojects detected bounding boxes directly into the ego-vehicle's local coordinate frame [Orange- Infra, Blue- Ego]. Validated system latency and spatial accuracy through live field testing at UC Riverside campus intersections.
Tech: PyTorch, MMDetection3D, ROS, LiDAR
LiDAR Differential Robot
ACM @ UC Riverside
Designed the perception and SLAM (Simultaneous Localization and Mapping) stack for an autonomous navigation robot, built on a differential-drive system and powered by ROS2 on a Raspberry Pi 5. To enable real-time environmental mapping of the campus engineering building, the SLAM system fuses data from a 2D LiDAR sensor, an IMU, wheel encoders, and GPS. For safe operation in dynamic environments, a YOLO-based obstacle classification model was integrated, achieving a 97.1% mAP@50 accuracy. By leveraging a priority-based A* path planning algorithm, the robot intelligently computes optimal routes to actively avoid obstacles.
Tech: ROS2, Raspberry Pi, LiDAR, YOLO, SLAM
MC Hammer
LA Hacks 2026
Engineered an offline, IDE-integrated AI pipeline on an ASUS Ascent GX10 supercomputer to autonomously resolve complex merge conflicts. Grounded in a 10,000+ commit dependency graph, the system executes a 3-stage local process: fix prediction, unbiased test validation, and refinement, to catch hidden downstream regressions before merging. By leveraging edge inference to eliminate cloud latency, this architecture accelerates conflict diagnosis and verified fix generation from over three minutes to just 45 seconds.
Tech: TypeScript, Python, FastAPI, NetworkX, Ollama
GitHub · Devpost
Real Estate Reality
RoseHack 2026 · 1st Place Overall
Built a Playwright-based scraping pipeline and React/TypeScript Chrome extension that analyzes Zillow listings in real time. Engineered concurrent geospatial scoring algorithms using OpenStreetMap (Nominatim/Overpass), the Haversine formula, and A* proximity search across six safety and convenience metrics to rank 37+ listings in under a minute, uncovering the safest homes hidden in search results. Won 1st Place Overall (40 teams) at RoseHack 2026.
Tech: React, TypeScript, Python, FastAPI, Playwright
GitHub · Devpost
Knight's Code
Cutie Hack 2025 · 2nd Place Overall
Architected a gamified learning platform for UCR's CS010 series that parses instructor-uploaded syllabi into structured AstraDB knowledge bases. Built a narrative-driven progression engine where students complete MCQs and live IDE debugging challenges to unlock inventory items that uncover a medieval mystery, increasing engagement through game mechanics.
Tech: Next.js, TypeScript, FastAPI, Gemini, AstraDB
GitHub · Devpost
Reel-To-Real
SB Hacks XII · TwelveLabs API Challenge Winner
Built and deployed a multi-source travel planning pipeline that cross-references 30+ travel sources into verified, ready-to-reserve recommendations, reducing itinerary planning time from ~10 hours to under 2 minutes. Implemented a custom Traveling Salesman Problem (TSP) routing algorithm that minimizes travel distance and cost by eliminating backtracking while respecting real-world time windows and geospatial constraints.
Tech: React, TypeScript, FastAPI, spaCy, TwelveLabs
GitHub · Devpost
ClassTrackr
VIBGYOR High School
In 11th grade, after observing teachers struggle with a school-wide transition from whiteboards to SmartBoards, I designed and deployed ClassTrackr. Built completely in Java, this classroom management app features a centralized database for real-time, school-wide usage tracking and an Over-the-Air (OTA) update system for seamless remote deployment. By automating attendance, assignment logging, and daily activity-summary emails to students, the application completely streamlined administrative workflows. Following a successful pilot, ClassTrackr organically achieved 85% adoption across 8th-12th grade classrooms within a month and remains in active use over two years later.
Tech: Java, Desktop UI, Automation
