Show HN: Resilient RAP – A self-healing pipeline for telemetry using embeddings

2 points

4 months ago

I spent some time building a prototype to handle schema drift in high velocity data. It is called the Resilient RAP Framework and it uses semantic embeddings to fix broken data streams on the fly.

I tested it by feeding it F1 telemetry and then switching to clinical ICU data to see if it could map different fields without manual help. It successfully mapped drift across domains, like matching kph to km/h in racing data and resolving pulse oximetry to oxygen saturation in a health context.

The repo hit about 600 clones on GitHub this week. I'm interested to see if anyone else is using agentic recovery for ingestion or if there are better ways to handle this at scale.

1 comment