Edge AI System Designer

Runs 8 sizing engines in parallel
Task type
Environment
Connectivity model
Number of cameras
Custom: cameras
Deployment preference
Resolution
Target FPS
Recording retention
Custom: days
Task type
Model complexity
Inference mode
Precision
Latency target (optional)
ms maximum
Power source
Power mode (Jetson only)
Max power (optional)
W maximum
Cooling
Budget range (platform cost)
Optimization goal
Analysing deployment…
No analysis yet
Configure your deployment on the left and click Analyze Deployment to generate a system architecture recommendation.
WHY THIS RECOMMENDATION
COMPUTE SUMMARY
ALTERNATIVES
Analysing alternatives…
DEPLOYMENT
TRADEOFFS
EXCLUDED PLATFORMS
Generate Deployment BOM →
// FAQ
What does the Edge AI System Designer do?

It takes your deployment description — camera count, model/task, FPS target, resolution, power source, and budget — and runs eight sizing engines in parallel to return a recommended hardware platform with compute, memory, power, storage, and network sizing, a bill of materials, a confidence label, and the risks that would break the design at scale.

How does it calculate the recommendation?

It normalizes your inputs into a canonical model, eliminates platforms that fail hard constraints (power, thermal, decode, memory), scores the survivors on a weighted set of dimensions, and validates real stream capacity against benchmark data. Confidence reflects how well-grounded the estimate is — proven benchmark versus interpolated — not marketing TOPS.

How is it different from the Hardware Selector?

The Hardware Selector answers "which platform fits?" fast — five inputs to a ranked shortlist. The System Designer goes further: it sizes the whole deployment (compute, storage, network, power, cost) and lets you stress-test it. Use the Selector to pick a platform, then the Designer to design the system around it.

Do I need an account, and is my data sent anywhere?

No account, and it is free. Your inputs and results stay in your browser; you can export the result as machine-readable JSON or copy a shareable URL that re-runs the same configuration.