GitHub
Stop losing data science context. Build knowledge graphs.
Start new thread
Rajiv Sambasivan

GitHub - Stop losing data science context. Build knowledge graphs.

by•
KMDS turns fragmented notebooks and data workflows into structured, searchable knowledge graphs. Scan repos using local LLMs, chat with your experimental history, and visually audit your data engineering artifacts—all 100% locally.

Add a comment

Replies

Best
Rajiv Sambasivan
Maker
📌
Hey Product Hunt community! 👋As data teams, we have all been there: a data scientist leaves the company, or you revisit a project after six months, and the research trail is completely cold. You are left staring at fragmented Jupyter notebooks, wondering why a specific modeling decision or data engineering choice was made.We built KMDS (Knowledge Management for Data Science) to solve this exact problem. It is an open-source, ontology-backed ecosystem designed to capture, organize, and reuse insights from your data science experiments.Key Features:Local Repo Scanning: Use local LLMs acting as specialized personas (Data Scientist, Tech Lead, Architect) to auto-generate structured knowledge graphs from your codebase.Interactive UI Workbench: Visually audit, explore, and safely edit your knowledge graphs without messing up raw files.Natural Language Ingestion: Simply describe your experimental insights in plain English, and KMDS maps them into the ontology.Semantic AI Search: Query your team's historical findings using local vector indices powered by Ollama.KMDS runs entirely on your local machine to keep your data and IP secure.Check out our GitHub repository, give it a spin, and let us know your thoughts! What features or integrations should we build next? We would love your feedback! 👇