RAG & LLM Agents | Data Science & ML | FastAPI & React
Computer engineering graduate (ESPRIM, 2024, data science option) and freelance AI Engineer since August 2025.
- AI: I build LLM applications end to end: agentic workflows with LangGraph, hybrid retrieval over Qdrant and FAISS, embedding-based intent matching, and evaluation sets (LangSmith, RAGAS) that catch regressions.
- Data science: foundation in scikit-learn, TensorFlow and MLflow, with leakage-aware validation.
- Backend & web: I serve AI features through FastAPI (REST and WebSockets) with PostgreSQL or MongoDB, and build React interfaces when a feature needs one.
AI & LLM Engineering
Data Science & ML
Backend & Web
Data & Infrastructure
AI Engineering Lab
Agentic RAG assistant: LangGraph tool routing over Gemini, hybrid Qdrant retrieval (dense + sparse with reciprocal-rank fusion), MCP integrations, streamed responses, PostgreSQL checkpoints, RAGAS evaluation and LangSmith tracing.
FastAPI LangGraph Gemini Qdrant PostgreSQL React
ResolveIQ
Triage prototype for NYC 311 heat and hot-water cases: a slow-resolution risk model validated on a chronological split with leakage masked (ROC-AUC 0.614 on one cohort), plus similar-case retrieval.
scikit-learn FastAPI LangGraph Qdrant Docker
Conversational AI Platform
Final-year project: a dialogue builder for intents, patterns and responses, and a chatbot that matches messages with Sentence Transformer embeddings and falls back to Gemma 2 2B via Ollama. Includes PDF question answering, speech features, feedback and analytics.
FastAPI MongoDB LangChain FAISS Ollama React
More: GenAI Code Reviewer (Gemini) · Assistive App for Visually Impaired Users (OCR, scene description, text-to-speech) · MERN Inventory Management (React frontend) · all repositories
- LLM Agents MOOC, UC Berkeley (2024)
- 365 Data Science: Build Chat Applications with OpenAI and LangChain (2024); Deep Learning with TensorFlow 2 (2023)
Open to AI Engineer / Applied AI roles. Reach me at [email protected].