a working catalogue
more in the lab →
reproducing the chat model training pipeline, end to end.
agentic retrieval over a corpus of support tickets
124M params, custom training loop
130 pods, 75GiB service, one config store