Ask in plain English
“Why is rank 42 stalling?” “Where is my MPI imbalance?” Seer’s AI interprets the question, then runs the right analysis to answer it.
AI-powered HPC Performance Analysis
A new way to understand your code. Seer launches targeted analysis pipelines to pinpoint performance issues in your app.
01/Product
Straight to the problem.
Most profilers hand you a mountain of traces. Seer points to the bottleneck and tells you how to fix it.
Targeted analysis pipelines
Every question is translated into a pipeline of novel, purpose-built analyses. You get the answer backed by evidence, visualizations, and annotations.
Vistas, not traces
A Vista is Seer’s high-level view of your application, with drill-down into functions, ranks, and calls on demand.
Shareable with your team
Send a Vista to a collaborator the way you’d send a link. Review each other’s runs, point to a hotspot, and keep the conversation on the data.
Zero-friction adoption
Our data collector captures your application's call tree and trace data without requiring any modifications to your source code. And our web-based interface means no new software to install.
Works with your stack
MPI, Kokkos, CUDA, OpenMP. Seer reads every source it can to build a full picture of your app.
02/How it works
From question to answer in three steps.
Seer closes the loop between collection, analysis, and visualization so you spend time fixing, not searching.
Collect
Attach Seer’s collector to your run. Minimal overhead, no source changes or recompilation required.
Ask
Describe what you want to know in plain English. Seer’s AI turns the question into a concrete analysis plan.
Explore
Answers land in a Vista: interactive visualizations you can drill into, annotate, and share with your team.
03/Vistas
Take in the view.
A Vista is your customizable analysis dashboard for a single run, composed of interactive visualizations built specifically for HPC data.
Overview first
See the whole run at a glance.
Start from the aggregate: execution timelines, node heatmaps, and communication maps sized for runs with thousands of ranks. No scrolling through complex traces to find the anomaly. We point you straight to it.
Drill down
Zoom straight to the call that matters.
Click into any hotspot to expand function-level calls, per-rank traces, and the supporting analyses. Annotations from Seer’s AI explain what it found and why it flagged it.
04/Collaborators wanted
Can your app run faster?
Of course it can! We’re looking for teams with real codes and real performance
problems.
Fill out the form below to get in touch with us.
Research groups
University and lab teams exploring parallel performance on new architectures.
Production HPC
Application owners running at scale who need answers faster than a traditional profiler gives them.
Facility staff
Center and system teams supporting many users across many stacks.
Seer is funded by the U.S. Department of Energy, Office of Science, Office of Advanced Scientific Computing Research, through the Small Business Innovation Research (SBIR) program. The WorkVisualizer prototype was developed under SBIR Phase I Award DE-SC-0024832.