Two-person team hits SOTA in GPU failure prediction with a $2 PINN algorithm
sharpeye_wnl · x · 2026-09-05
Developer @prb5h recaps one year of the 'lock-in' project with @chastronomic: starting from BCIs, they built a near-SOTA EEG denoising algorithm at a fraction of usual cost, then — inspired by GPU failures discussed in Meta's blog — treated the problem as signal processing. With @nihipap joining and a move to Bangalore (lossfunk), they reached SOTA in GPU failure prediction using a PINN-based algorithm built for $2, secured $52k in grants (Emergent Ventures, AI Grants India, 3F VC), entered NVIDIA's Inception Program, and got press coverage.
More from Infra
- Reef launches inference-native infra that serves self-improving agents without downtime — pliang279 · 2026-09-05
- Google DeepMind Publishes Free Book on Scaling LLMs Across TPUs and GPUs — goyal__pramod · 2026-09-05
- From 1D to 2D int8 kernels: a hands-on GPU internals learning path — goyal__pramod · 2026-09-05
- Prime Super Flash MoE: 1.2x BF16 and 1.6x MXFP8 speedups over upstream on B200 — retr0jirachi · 2026-09-05
- Prime Intellect cuts GLM-5.2 RL weight transfer from 86s to 4s with NIXL and ModelExpress — samsja19 · 2026-09-05
- How to Estimate tokens/sec on Your Hardware: The VRAM Bandwidth Formula — Pyrolistical · 2026-09-05