Datology releases Zephon, a deterministic on-the-fly dataloader born from MosaicML Streaming's legacy
josh_wills · x · 2026-10-10
Datology AI released Zephon, a dataloader that processes data on the fly and deterministically, targeting two recurring pains at scale: evals differing between 8- and 16-GPU runs of the same job, and having to re-materialize datasets after every pipeline change.
The founder — an ex-MosaicML person — calls MosaicML's Streaming dataloader one of the most underrated pieces of ML infrastructure ever, and says building its torchbearer was a founding goal of Datology: well-designed dataloaders enable real-time curricula, arbitrary mixing schedules, and seamless pre/post-training transitions, then make those experiments productionizable. Zephon already powers thousands of internal data experiments.
More from Infra
- Starship Could Cut Cost to Orbit 100x to ~$185K Per Ton, Unlocking New Businesses — claud_fuen · 2026-10-10
- DuckDB v2.0 CLI agent mode cuts agent-read tokens by 59% on TPC-H benchmarks — josh_wills · 2026-10-10
- Tsinghua's TokenRouter: Token-Level LLM Routing Hits Up to 64.15X Serving Throughput — rohanpaul_ai · 2026-10-10
- Meta Muse Auto-Routes to OpenRouter Free Models for Zero-Cost Long Tasks — sven_ai · 2026-10-10
- Nvidia CEO's son-in-law becomes VP as Apple cuts iPhone 18 Pro orders 15%+ — 创业邦 · 2026-10-10
- Ex-NVIDIA GPU veterans build Physical AI chip claiming 5x performance at 80% lower cost — 创业邦 · 2026-10-10