Join the conversation

Join the community of Machine Learners and AI enthusiasts.

Sign Up
karstenskytย 
posted an update 1 day ago
Post
72
โšฝ We've been building ๐š•๐šž๐šก๐šž๐š›๐šข-๐š•๐šŠ๐š”๐šŽ๐š‘๐š˜๐šž๐šœ๐šŽโ€”an open-source soccer analytics platformโ€”and we're using the Hugging Face Hub as our public distribution layer.

"Luxury! We used to dream of serverless!" We replaced a traditional 6-service AWS pipeline with a unified lakehouse architecture. While Databricks handles our backend Medallion architecture (processing ๐Ÿฏ๐Ÿด๐— + ๐˜๐—ฟ๐—ฎ๐—ฐ๐—ธ๐—ถ๐—ป๐—ด ๐—ณ๐—ฟ๐—ฎ๐—บ๐—ฒ๐˜€ from 5 vendors), we rely entirely on the HF ecosystem for public access and specific compute workloads.

I just updated our org page with a full architecture breakdown showing how we integrate ๐—›๐—™ ๐— ๐—ผ๐—ฑ๐—ฒ๐—น๐˜€, ๐——๐—ฎ๐˜๐—ฎ๐˜€๐—ฒ๐˜๐˜€, ๐—š๐—ฟ๐—ฎ๐—ฑ๐—ถ๐—ผ ๐—ฆ๐—ฝ๐—ฎ๐—ฐ๐—ฒ๐˜€, and use ๐—›๐—™ ๐—๐—ผ๐—ฏ๐˜€ for serverless Expected Threat (xT) computation with a bidirectional Databricks sync.

Check out the full architecture details here: ๐Ÿ‘‡

luxury-lakehouse