Graduation Thesis
Completed
DermaLens
AI-Powered Dermatological Assistant
- Engineered CNN with TensorFlow achieving 90%+ skin disease classification accuracy
- Deployed via TensorFlow Lite for mobile inference
- Built Flutter mobile app integrating AI model and Google Maps API for dermatologist navigation
TensorFlow
CNN
Flutter
Google Maps API
TF Lite
Personal Project
Active Dev
Argus
Autonomous AI research assistant powered by a multi-agent orchestration pipeline
- Designed a multi-agent system using LangGraph with specialized roles: Planner, Researcher, Critic, and Writer — each agent handling a distinct phase of the research pipeline
- Integrated Tavily Search API for real-time web retrieval; Critic agent validates source quality and triggers re-research when confidence is low
- Implemented Redis-backed session state for mid-task persistence and context continuity across agent nodes
- Exposed async FastAPI endpoints with WebSocket support for real-time agent step streaming to the frontend
- Benchmarked at <45s end-to-end report generation across 5-source research tasks
LangGraph
FastAPI
Redis
WebSockets
Tavily API
Personal Project
Completed
Nexus
Smart Documentation Buddy
- Production-grade RAG assistant enabling semantic search and Q&A over technical PDFs
- Conversation memory with LLM-based question reformulation for context-aware follow-ups
- React + Vite frontend with drag & drop ingestion, interactive querying, and source-cited responses
LangChain
FastAPI
Ollama
React
Vite
