Zymtrace, a developer infrastructure startup building continuous optimization tools for GPU and AI workloads, announced $12.2M in total funding including an $8.5M seed round led by Venture Guides. Mango Capital and Fly Ventures increased participation following earlier pre-seed backing, with 6 Degrees Capital and Concept Ventures joining the round. Funding will accelerate development of a platform designed to improve performance and utilization of enterprise GPU clusters.
GPU Optimization Platform
Zymtrace develops a continuous profiling platform that analyzes GPU and CPU workloads across distributed infrastructure. The system identifies performance bottlenecks inside AI training and inference pipelines by correlating execution data from CPU code paths down to CUDA kernels and GPU instruction-level stalls. The platform operates with zero code modifications by using low-overhead instrumentation based on eBPF technology.
Many enterprise GPU clusters currently operate at roughly 35–40% utilization due to inefficient code execution and limited visibility into performance bottlenecks. Zymtrace provides cluster-wide insights that enable engineering teams to diagnose slow workloads and improve performance without purchasing additional hardware.
AI-Driven Optimization
Zymtrace also introduces Profile-Guided AI Optimization, exposing profiling data directly to AI agents through MCP. The system allows AI agents to analyze performance data and generate code fixes automatically, enabling faster remediation of performance issues within training and inference pipelines.
Product Impact and Growth
Organizations using Zymtrace report improvements in inference latency and throughput by optimizing existing infrastructure rather than expanding GPU capacity. The company plans to develop autonomous optimization capabilities that detect GPU bottlenecks and automatically generate pull requests with code fixes.
Zymtrace operates as a distributed team and continues expanding global engineering and go-to-market operations to support growing demand for AI infrastructure optimization.