AI Engineer focused on building production-grade machine learning systems — from deep learning models to deployable APIs and evaluation pipelines.

I primarily work with PyTorch, and my interests include computer vision, reinforcement learning, and LLM-based systems.

🌐 Portfolio: https://sanskarmodi8.github.io

Hybrid retrieval system combining dense embeddings, BM25, and graph-based concept expansion.
FastAPI backend with document ingestion, citation-aware responses, and retrieval evaluation.
👉 https://github.com/sanskarmodi8/Atlas-RAG

PPO-based Bitcoin trading agent achieving 204% return vs 124% buy-and-hold baseline.
MLflow + DVC used for experiment tracking and reproducibility.
👉 https://github.com/sanskarmodi8/MarketMind

Multi-modal MRI segmentation pipeline using Swin UNETR (PyTorch, MONAI).
Includes reproducible training, evaluation, and 3D visualization.
👉 https://github.com/sanskarmodi8/brain_tumor_segmentation

Video-based deepfake detection system with visual explainability.
Deployed on Hugging Face Spaces.
👉 https://github.com/sanskarmodi8/DeepDetect

Languages: Python, Java
ML / DL: PyTorch, Scikit-learn, OpenCV
GenAI: Hugging Face, LangChain
Backend: FastAPI, Docker, Linux
MLOps: MLflow, DVC
Prototyping: Streamlit, Gradio

• LinkedIn: https://www.linkedin.com/in/sanskar-modi-ba53a2267
• GitHub: https://github.com/sanskarmodi8
• Email: mailto:sansyprog8@gmail.com
• Kaggle: https://www.kaggle.com/sanskarmodi8