Last year, I (Alex) was onboarding at a biotech company in the R&D phase applying ML. As you can imagine, when a team of scientists from non CS-background come together, the code gets pretty messy and full of domain-specific quirks. During my time there, I asked a lot of questions, yet when my internship came to an end, I couldn't explain to a new hire in under thirty minutes what "preprocessor_KLN_v3.py" was doing. I had gotten used to the mess.

Our diagram generation combines static analysis and LLM agents to scale to bigger projects, that otherwise could not be processed by pure LLMs, and to reduce hallucinations. We create "abstract" components you can click to go to a lower level, along with short summaries for every component.

You can explore sample maps here: https://github.com/CodeBoarding/GeneratedOnBoardings and generate new ones (Python only atm.) here: https://codeboarding.org/demo. In two weeks, we’ll open up a small batch of free hosted licenses for a VSCode extension - post below if you’d like one.

We’d love to hear your feedback on the demo to improve onboarding. Cheers!