I am a Computer Science and Engineering major at UC Merced, specializing in full-stack development, mobile development, and machine learning. Currently serving as a Software Engineering Intern at Arrowz, a mental health startup, where I develop cross-platform mobile applications using Flutter/Dart. As a Project Manager for both the Machine Learning Club and Data Science Society Club, I lead teams building AI-powered tools like Code Companion and Research Copilot. I'm passionate about building AI-powered applications and mobile solutions that solve real-world problems, from containerized web apps with GPT-4 integration to real-time computer vision systems. My experience spans modern web technologies like React and TypeScript, mobile development with Flutter, backend systems with Python and Flask, and cutting-edge machine learning frameworks including PyTorch and OpenCV. I excel at creating innovative solutions that bridge AI research with practical, user-friendly applications, while maintaining focus on performance optimization and seamless user experiences.
Shipped cross-platform Flutter/Dart app for emotion tracking across 25+ screens; modular UI components cut duplication 40%. Standardized state management and responsive layouts; improved build stability and developer velocity across the team.
Scoped and led Code Companion, an AI-powered Python debugging tool that ingests GitHub repos, runs flake8/mypy, and suggests LLM-generated fixes with explanations. Defined MVP, metrics, and phased roadmap for Next.js + FastAPI + pgvector stack, coordinating backend and frontend milestones with the team.
Managed a RAG-based research assistant that ingests PDFs/arXiv links, runs hybrid BM25 + embedding search, and answers questions with citation-grounded quotes. Designed system architecture (Next.js, FastAPI, Postgres + pgvector) and evaluation plan focusing on answer quality, latency, and pilot user feedback.
A–Z ASL recognition with emotion-aware voice chat; ONNX Runtime enables client-side inference at 90 FPS. Trained CNN in PyTorch with augmentation and mixed precision; exported to ONNX with dynamic axes. Built Next.js UI (App Router, Tailwind) with accessible components and graceful fallbacks for camera/mic permissions.
Fine-tuned ResNet-18 on Cats vs Dogs to 98.6% validation accuracy; used cyclic LR and label smoothing. 30 FPS webcam inference via OpenCV pipeline (resize/normalize, softmax thresholding, class overlay). Reproducible training scripts and Jupyter notebooks; automated environment setup with requirements.txt.
Containerized full-stack app that ingests PDFs and returns structured, role-aligned feedback with job matches. RESTful Flask service: PDF parsing via pdfminer, input validation, rate limiting, and CORS-secured endpoints. Realtime updates to Firebase; typed React UI that highlights skills gaps and exports notes to Markdown.
Natural-language trip planning using Gemini + Maps Grounding; parses intent and constraints to build candidate routes. Dynamic stop matching across 7+ routes using Haversine distance and snapping to road geometry for realism. Secure API routes for geocoding and distance matrix; deployed on Vercel with interactive map and ETA breakdowns.
Created the frontend of EdTok, a platform designed to transform student learning through short-form content and quizzes. EdTok allows teachers to upload educational videos organized by classes, subjects, and chapters, followed by quizzes to test student understanding.
Building a free website where users can interact with multiple LLMs, including GPT-4, Claude 3.0, Meta AI, and Gemini. User data is securely stored via Firebase. The platform is funded through Google Ads.
I'm always interested in new opportunities and collaborations. Feel free to reach out if you'd like to work together!