# Vaibhav Gupta — Junior AI/ML Engineer

> Junior AI/ML Engineer at Tracky (Drstore Healthcare Services India Pvt. Ltd.), Mumbai, India. First engineer on Tracky's AI platform, converted to full-time after architecting its LangGraph-based health chatbot. B.Tech Computer Science & Engineering (Cybersecurity), Thakur College of Engineering and Technology (TCET), Mumbai — CGPA 9.39/10.

This file is a machine-readable summary for AI agents and LLM-based tools. Full site: https://vaibhavgupta-dev.vercel.app/

## Experience

- **Junior AI/ML Engineer**, Drstore Healthcare Services (Tracky) — Jul 2026–Present (Full-time). Rewrote health chatbot from a single LangGraph agent into a router → orchestrator → worker graph backed by 7 independent MCP servers, cutting tool bloat and isolating new device support to a single change. Designed concurrency-safe LangGraph state with custom reducers, runtime-planned fanout, and empirically-tuned re-planning limits. Led a codebase-wide optimization audit (pooled psycopg3 singleton, 5-layer module convention). Built the privacy/auth layer (HMAC user-ID hashing, JWT/JWKS gateway auth). Shipped OAuth integrations for Whoop, Garmin, and Fitbit. Wrote an automated LLM-judge evaluation harness for health-answer accuracy.
- **AI/ML Intern**, Drstore Healthcare Services (Tracky) — Aug 2025–Jun 2026. Worked on real-world biometric datasets (CGM, HRV, sleep, stress, recovery). Built AI/ML data pipelines and analytical models turning physiological data into wellness insights.

## Skills

- **AI/ML**: Feature Engineering, Model Optimization, ML Pipelines, NLP, Neural Networks & Deep Learning, LLMs & Transformers, RAG, Agentic AI Systems
- **Languages**: Python, C++, Java, Go
- **Frameworks**: LangChain, LangGraph, LangSmith, PyTorch, TensorFlow, Scikit-learn, Hugging Face Transformers, FastAPI, Flask, Streamlit
- **Infra/Tools**: FAISS, Pinecone, Weaviate, Ollama, OpenAI API, Google Gemini, Groq, Git/GitHub, Jupyter, Google Colab, Kaggle

## Projects

- **Deep Research Agentic-AI System** — LangGraph multi-agent system (Researcher/Drafter/Enhancer) for scholarly research automation, Groq-powered, LangSmith observability. Code: https://github.com/gupta-v/Deep-Research-Agentic-AI
- **Reddit Persona Builder** — Multi-agent pipeline generating cited user personas from public Reddit activity, LangChain + Groq. Code: https://github.com/gupta-v/reddit-persona-builder
- **KIRA — AI Learning Assistant** — RAG assistant that routes between PDF-embedding retrieval (FAISS) and live web search, with voice I/O. Code: https://github.com/gupta-v/KIRA
- **AI Video Analyzer & Chat Agent** — Video understanding + conversational chat agent using Gemini 1.5 Flash, LangChain, Agno. Code: https://github.com/gupta-v/ai-video-analyzer
- **Movie Recommendation System** — Content-based recommender using TF-IDF vectorization + Pinecone vector search. Demo: https://movie-recommender-zeic.onrender.com/ · Code: https://github.com/gupta-v/movie-recommendation-system
- **Valorant Performance Predictor** — Gradient Boosting Regressor (R²=0.8945) predicting player win probability from gameplay stats. Demo: https://valorant-performance-prediction-gupta-v.streamlit.app/ · Code: https://github.com/gupta-v/valorant-performance-predictor
- **Eva — Desktop Assistant** — Voice-driven automation assistant, Gemini API + Flask, speech recognition/TTS, fuzzy command matching. Code: https://github.com/gupta-v/Eva

## Contact

- Email: vaibhavdgupta18@gmail.com
- GitHub: https://github.com/gupta-v
- LinkedIn: https://www.linkedin.com/in/gupta-v18/