class DebdeepGhosh:
def __init__(self):
self.role = "AI / ML Engineer"
self.focus = ["Computer Vision", "RAG & LLM Systems", "Fraud Detection"]
self.stack = ["Python", "FastAPI", "Streamlit", "OpenCV", "YOLOv8"]
self.currently = "Building end-to-end ML systems, from model to API to UI"
def philosophy(self):
return "Ship things that work end-to-end, not just notebooks."
- 🔭 Building practical, end-to-end AI systems — fraud detection, retrieval-augmented research tools, and 3D perception
- 🧠 Deep interest in applied computer vision (stereo cameras, point clouds, object localization)
- 🛠️ Comfortable across the stack: data → model → API → interface
- 📈 Always shipping — check the pinned repos below
| Repo | What it does |
|---|---|
| Insurance_fraud_detection | End-to-end ML app predicting fraudulent insurance claims with a Random Forest Classifier — FastAPI + Streamlit + SQLAlchemy |
| AI_Research_Assistant | Retrieval-Augmented Generation (RAG) research assistant built with Streamlit |
| ZED-3D-Localization | Robust 3D object localization using a ZED stereo camera, YOLOv8m, and point-cloud processing with manual ROI fallback |
| Docling_pdfextract | Extracts text, tables, and images from PDFs, structured by page, using Docling-style parsing in Python |
