Zymtrace raises $12.2M and hires ML engineers to build agents that fix GPU bottlenecks
cfregly · x · 2026-10-16
Zymtrace announced a $12.2M raise and is hiring a Senior ML Engineer (fully remote) to build AI agents that autonomously diagnose and fix GPU performance bottlenecks.
- Positioning: a distributed AI infra optimization platform claiming 60–65% of GPU spend is wasted on idle cycles and inefficient workloads, while existing tools treat GPUs as black boxes
- Approach: traces from PyTorch/JAX code through CUDA dispatch down to individual GPU instructions and stall reasons, correlated back to the triggering CPU code — zero code changes required
- Team pedigree: founders open-sourced the eBPF continuous CPU profiler now in OpenTelemetry, used in production by Google, IBM, Datadog, and Grafana
- Role scope: Python, Rust, C++, GPU kernels, and hardware primitives across the full execution path
More from Venture
- Musk endorses idea of an agent-first, US-only Amazon competitor — elonmusk · 2026-10-16
- REAX SN92 targets the AI decision economy: small models making millions of micro-decisions daily — markjeffrey · 2026-10-16
- Bittensor SN62 Ridges ties agent upload fees to demand: +15% per upload, $0.29 inference budget — bittingthembits · 2026-10-16
- Founder: GTM now burns nearly as much as engineering in the AI era — thedealdirector · 2026-10-16
- Creative Strategies launches CS Forecast Stack: 19 linked models covering AI's full economic chain — BenBajarin · 2026-10-16
- CNN signs multi-year AI licensing partnership with Google for generative AI experiments — gaganghotra_ · 2026-10-16