My Projects
Here are some of the projects I've worked on. Each one represents a unique challenge and learning experience in my journey as a developer.
Multi-Tool MCP
Built an MCP server with Next.js and TypeScript using the official template. Integrated Gmail, Google Calendar, and Google Drive for meeting scheduling, email context, and documentation. Used OAuth2 to authenticate API calls while preserving user identity and permissions. Coordinated multi-tool workflows via MCP protocol with stateless, token-bound server interactions.
Skills Used:
Automated Insurance Claims
Automated extraction and population of clinical data from complex referral PDFs for prior authorization. Parsed 30–50 page packets using Fitz and Mistral OCR, handling poor scan quality and inconsistent formats. Populated insurer-specific forms with key patient and treatment data, including conditional logic. Used Gemini API to interpret and structure clinical data for context-aware form completion.
Skills Used:
Multimodal Video Analysis App
Built a web app using Next.js and Tailwind that lets users upload YouTube links and chat with video content. Generated section breakdowns and timestamped citations using YouTube-transcript and Gemini API for contextual navigation. Enabled natural language queries to search video frames by visual content via CLIP/OpenCLIP and PyTorch. Orchestrated backend with FastAPI for video processing and Redis for performance optimization.
Skills Used:
Neural Nocturne
Implemented a neural network in Python with TensorFlow and Keras for accurate music genre prediction. Utilized features such as spectral characteristics and rhythm patterns, reaching a 93% prediction accuracy. Enhanced machine learning proficiency and strengthened collaborative and problem-solving skills.