Welcome to my data portfolio! Here, I document a summary of my projects in the data field.
| Project Link | Tools | Description |
|---|---|---|
| Real-Time Expression Detection | OpenCV, Keras, TensorFlow | CNN-based real-time facial expression recognition from video streams. |
| Project Link | Tools | Description |
|---|---|---|
| Hospital LOS Prediction | scikit-learn, pandas | ML model predicting hospital length of stay from 100k records. XGBoost achieved 0.3032-day MAE, beating baselines. |
| Project Link | Tools | Description |
|---|---|---|
| PDF RAG Chatbot | LangChain, ChromaDB, Ollama, Streamlit, Docker | Fully local RAG pipeline to chat with multiple PDFs. Chunks documents into embeddings, retrieves semantically similar context, and generates grounded answers with page-level source citations. No API costs — runs entirely on-device via Ollama. |
| Healthcare Research Agent | LangChain, OpenAI API, Streamlit, Docker | 4-Agent AI system searching PubMed to create medical research reports. |
| CiteIQ | LangGraph, ChromaDB, LangSmith, HuggingFace Spaces | Production agentic RAG system with LangGraph multi-agent orchestration, hybrid retrieval, and LLM-as-judge self-evaluation (faithfulness + relevancy scoring). Processes clinical medical literature with 1.0 faithfulness scores. Deployed live at harshxth-citeiq.hf.space. |
| Project Link | Tools | Description |
|---|---|---|
| Hospital Readmission Pipeline | Apache Airflow, BigQuery, GCS, Python, SQL | End-to-end GCP data pipeline for hospital readmission analysis and revenue insights. Implements Bronze → Silver → Gold medallion architecture in BigQuery, orchestrated via Cloud Composer. |
(PowerBI, Tableau, SQL analysis will push soon)
