ARTIFACT
Dockerized microservice with scripts
raxhacks // ai-infra-swe
Raxhacks AI Infra SWE Learning Roadmap
A 4-week, hands-on path from Linux and containers to production inference infrastructure. Check off each day, log your notes, and ship the weekly artifact as you go.
ARTIFACT
Quantized model benchmark on local Mac
BENCHMARKlocal mac - 18 GB
| precision | memory | result |
|---|---|---|
| FP16 | 18 GB (maxed) | UI locked, swap choked - KV cache set to 3.6 GB caused the issue. |
| INT4 | 14 GB / 18 GB | Much better performance with headroom to spare. |
ARTIFACT
Containerized inference service with API
ARTIFACT
End-to-end AI infrastructure capstone