AI Engineer • Full-Stack Developer • Problem Solver ---

Multi-Agent Workspace Generation via Stateful Graph Orchestration and AST Validation

• Stateful Multi-Agent Workflow: Coordinated a stateful Directed Acyclic Graph (DAG) pipeline using LangGraph and LangChain to manage context across four specialized concurrent agents (Research, Feasibility, Architecture, Review) without context drift.
• 5-Pass AST Static Validation: Engineered an Abstract Syntax Tree (AST) static analysis engine using a custom SymbolAndUndefinedVisitor to verify scoping chains, isolate unresolved global name errors, and check cross-file module contract compliance before filesystem serialization.
• Autonomous Self-Correction: Implemented a mathematical grading quality gate Score = 10.0 - ({errors} x 1.5) - ({warnings} x 0.3) that logs failure traces and automatically triggers up to 3 self-correction retry loops.
• Human-in-the-Loop Interrupts: Engineered bidirectional WebSocket channels to statefully suspend execution gates via LangGraph interrupt primitives, allowing developers to review blueprints via React Flow, audit live terminal logs, and inject real-time steering feedback.
• Eager Sync & Previews: Bound browser updates to a reactive Zustand store and cached React Query layers for eager keystroke synchronization, while the ASGI backend dynamically allocates ports to spawn isolated application live previews.
• Stack: FastAPI, LangGraph, Next.js 16 (React 19), ChromaDB, WebSockets, Tailwind CSS 4, React Flow.

Full-Stack Predictive Dashboard for Explainable Risk Modeling

• Meta-Dynamic Ensemble: Engineered a hybrid model fusing XGBoost (spatial severity) and LSTM (temporal volume) forecasting across 20,000 national records.
• Explainable AI (XAI): Integrated an interpretable logic layer to surface primary risk drivers, ensuring transparency for tactical decision-making.
• Conversational Intelligence: Built a RAG-powered assistant using Gemma 3 and ChromaDB for natural language interrogation of vectorized datasets.
• Stack: FastAPI, Next.js 14, TensorFlow, XGBoost, Leaflet.js.

• Architecture: Hybrid RAG merging FAISS Vector Search with live Web Search.
• Engineering: Built a non-blocking pipeline using Celery & Redis for async PDF processing.
• UI/UX: Custom PDF Viewer with bi-directional interactive citations and dynamic model switching (Gemini Flash/Pro).

Research Internship

• Multimodal AI: Fine-tuned LLaVA for technical multimodal tasks, improving accuracy in processing airworthiness and defense document intelligence.
• Scalable R&D: Authored deployment-ready documentation and optimized pipelines for high-security, defense-grade datasets.

• AI/ML/Orchestration: PyTorch, TensorFlow, XGBoost, Scikit-learn, LangGraph, LangChain, Ollama, HuggingFace.
• Backend: Python, FastAPI, Flask, PostgreSQL, Redis, Celery, SQLAlchemy, aiosqlite, Docker.
• Frontend: Next.js, React, React Flow, Monaco Editor, Zustand, React Query, Tailwind CSS 4, Recharts, Leaflet.js.
• Databases: SQL, ChromaDB, FAISS.

Building things that make people’s lives easier — one project at a time.