Universal and Reproducible Compute
summary
Bacalhau implements Compute over Data by running containers, WASM, or arbitrary binaries where the data lives, assigning a CID per job and per input, and producing cryptographic lineage and audit logs so every transformation is verifiable. It supports data-aware scheduling, local mounts, secure enclaves for gated execution, and runs from cloud to Raspberry Pi or phone. Used in scientific and academic cases, it claims to cut data movement and reduce regulatory exposure tied to roughly $250 billion in global fines.
AI-written summary · about
chapters (9)
- 0:00Introduction and delay
- 0:45What is universal reproducible compute
- 3:13Problems with current data handling
- 5:41Compute over data concept
- 7:47Introducing Baka Yaa project
- 9:13How Baka Yaa works demo
- 10:17Collaboration without data movement
- 13:50Security and audit trail
- 14:11Baka Yaa’s mission and community
related talks
24m🛠 Making Data Useful with Decentralised Compute
37mUniversal AI: Execute Your Models Where Your Data (And Users) Are
36mAn almost perfect match: cloud native and scientific computing
30mProject Caerus: Compute Storage Coordination
35mState of Containers and the Convergence of Hyperscale, BigData and HPC
33mBuilding Reproducible ML Processes with an Open Source Stack
41mScalable Multi-Language Data Analysis on Beam: The Cuneiform Experience
25mBuilding AI Workflows: From Local Experiments To Serving Users