Applied AI and Machine Learning Engineer building evaluated AI systems, ML pipelines, and developer-facing tools.
- Generative AI Intern at TransOrg Analytics (Pickl.AI) × LPU, where I led the technical implementation of an evaluated Text-to-SQL system.
- Creator of
vigil-drift, an open-source Python package for unsupervised concept-drift detection and feature attribution. - B.Tech (Hons.) CSE — Data Science and Data Engineering from Lovely Professional University, CGPA 8.38/10.
| Project | What I built | Evidence |
|---|---|---|
| F1InsightAI | Nine-node LangGraph pipeline for natural-language querying across 700,000+ Formula 1 records in TiDB Cloud | 83.3% first-attempt SQL accuracy (15/18 SQL queries); schema-retrieval MRR improved from 0.12 to 0.67 across three evaluation iterations · Live demo |
| Vigil | Python package for detecting and attributing distribution shifts without labels | 93.3% drift precision and 100% novel-class recall in the documented NSL-KDD experiment · PyPI · Docs |
| CTR Predictor | Large-scale tabular ML pipeline and scoring application for Criteo display-ad data | XGBoost test AUC 0.9067 versus LightGBM 0.9024 across 10 million records · Live app |
| P&ID Intelligence | Prototype that combines object detection, OCR, graph mapping, and validation to produce MTO records | OCR processing reduced from approximately 360 seconds to 7 seconds in project tests |
- Applied AI: LangGraph, LangChain, RAG, FAISS, sentence-transformers, LLM APIs
- Machine learning: PyTorch, scikit-learn, XGBoost, LightGBM, Optuna
- ML engineering: FastAPI, Flask, Docker, GitHub Actions, Kafka, Airflow, MLflow
- Data and vision: SQL, TiDB Cloud, pandas, OpenCV, YOLOv8, Tesseract OCR, NetworkX
I am open to entry-level Applied AI, Machine Learning, and Generative AI roles in India or remote teams. Contact: venkateswarsahu000@gmail.com.