← All organizations

Continuous profiling and performance optimization

Polar Signals

Polar Signals builds Polar Signals Cloud, a continuous profiler that helps engineering teams understand where production software spends CPU time, allocates memory and uses GPUs. Its eBPF agent collects performance data without application code changes, while flame graphs and profile comparisons help developers locate bottlenecks and regressions. An MCP integration gives AI coding agents access to profiling data, connecting optimization suggestions to actual execution. GPU profiling covers NVIDIA CUDA workloads, including training and inference.

Founded in Berlin in 2020 by founder and CEO Frederic Branczyk, Polar Signals grew from engineering work in the Prometheus ecosystem. It created Parca, an open-source profiling storage and query engine with an eBPF profiler. Its GPU engineering extends this approach to instruction-level CUDA sampling and explanations of why GPU execution stalls. Short, adaptive sampling windows and backend source mapping reduce the overhead of collecting that detail in production.

Vercel selected Polar Signals Cloud for its edge platform, as the company announced in 2024 alongside $6.8 million in additional seed funding, led by Spark Capital, bringing total capital raised to $10.8 million. Dash0 completed its acquisition of Polar Signals in August 2026; financial terms were undisclosed. Polar Signals Cloud and support continue, and Parca remains maintained and open source. Native integration into Dash0’s platform is planned.

www.polarsignals.com

1 talk

Newest first

1 speaker at AIE

Affiliations reflect their AIE appearances, not necessarily current employment.

Messages from the stage

Kernel timing and profiler deployment

Loibl demonstrated measuring CUDA kernel execution time and deploying the profiler through systemd, Docker, or Kubernetes.

Affiliations reflect each recorded session, not necessarily current employment.

Company sources · checked 2026-08-28