Test it. Tune it. Take it home.
The EdgeLab Platform

Test it. Tune it. Take it home.

EdgeLab isn't a benchmark tool.

It's structured, hands-on access to real edge devices: validate your model, learn what to improve, get hands-on with the hardware you're considering, then take the exact setup home.

The loop at a glance

Four stages. One loop.

Move through all four, or jump straight to the one you need.

The EdgeLab loop — Benchmark, Learn, Develop, Deploy
Benchmark

See how your workload actually behaves on real hardware.

Upload your model in the format you already use. Pick one device or compare several. EdgeLab runs your workload live on real edge hardware and shows you what happens.

You get the numbers that actually matter: latency, throughput, memory and resource utilisation, and power draw, tracked over time rather than a single snapshot. Understand how different hardware, software and configuration choices influence workload behaviour.

SUPPORTED FORMATS

ONNX
PyTorch
TensorFlow · coming soon

METRICS

Latency
Throughput
Resource utilisation (CPU / GPU / RAM)
Efficiency
Power draw
EdgeLab benchmark dashboard
Learn

Compare. Improve. Repeat.

EdgeLab version comparison view

Your first measurement is a starting point, not a conclusion. Adjust your model, your configuration, or your target device, and run it again. EdgeLab keeps every version side by side, so you can see exactly what changed.

Compare versions directly: which change actually improved latency, which one cost you accuracy, which configuration hit the best balance. Same metrics as in benchmark, now tracked across iterations instead of a single run.

WHAT YOU COMPARE

Versions
Configurations
Devices
Models

WHAT YOU LEARN

Trends over iterations
Metric deltas
Side-by-side comparison
Develop

Direct, hands-on access to the hardware.

Once you know which device fits, book time on it directly. No shared queue, no simulated environments. You get a dedicated session on the actual hardware, to understand how it really behaves.

Get shell access, run your own tools and push the device the way you would in production. It's the closest thing to having the hardware on your own desk. But without the need to buy it first.

Interactive EdgeLab device session
Deploy

Take the exact setup home.

EdgeLab deployment artifacts

Once you've found the setup that works, EdgeLab packages it up: the container image, the OS, and the exact configuration you tested and tuned. No rebuilding, no guessing which driver version you had: it's the same environment, ready to deploy.

Run it on your own hardware and it behaves the same way it did in EdgeLab. You bring the workload, we help you get it right, and you walk away with something that's fully yours.

WHAT YOU GET

Configuration
Kairos OS Image
Container Image
imec research campus building

Built by researchers. Not hardware vendors.

EdgeLab is built by imec, the same independent research institute that's worked at the edge of hardware and software for over 40 years. We don't sell chips, and we don't get paid to make one device look better than another. When a result favours one vendor over another, it's because the hardware earned it.

Every experiment runs on real hardware under consistent conditions, ensuring that performance insights are reproducible and trustworthy. The goal is not just to produce a number, but to understand workload behaviour, expose bottlenecks and support better deployment decisions. Re-run the same workload and you'll arrive at the same evidence.

Ready to test?

EdgeLab is currently in beta and available to a limited group of partners as we refine the platform. Tell us about your workload or hardware questions, and we'll get back to you within a few business days.

edgelab@imec.be