EdgeOpt
Prepare models for constrained devices and compare the result with the baseline.
- Index
- 33 / 39
- Division
- Computer vision & applied AI
- Stage
- Model optimisation for devices
- Built by
- Athreix, Pune
EdgeOpt is a model optimisation for devices from the Athreix Computer vision & applied AI division. Prepare models for constrained devices and compare the result with the baseline.
The problem
EdgeOpt was built for edge AI and embedded devices. Prepare models for constrained devices and compare the result with the baseline. The starting point was the work as it already happens: the people, the records and the decisions that have to be made every day, and the places where that work is slow, scattered or hard to check.
What Athreix built
Athreix built EdgeOpt as one connected system. It belongs to the Athreix computer vision and applied AI division, which builds inspection, perception and voice systems that have to work under real lighting, real noise and real throughput. EdgeOpt is a model optimisation for devices.
How EdgeOpt works
- 01Review the model
- 02Model compression + device conversion
- 03Compare + export
- Review the model: the system starts here, taking in the real inputs the team already has rather than asking them to change how they work.
- Model compression + device conversion: an intermediate step where the AI does the heavy work and keeps a trace of how it got there.
- Compare + export: the final step produces something a person can review, approve and act on, with the evidence kept alongside it.
What you can get
- 01Smaller model candidates
- 02Comparable device runtimes
- 03Visible accuracy trade-offs
With EdgeOpt, teams get smaller model candidates, comparable device runtimes and visible accuracy trade-offs. Every result keeps a human in control: a person can see what the system decided, why, and override it.
Who it is for
EdgeOpt suits edge AI and embedded devices that want AI inside the workflow they already run, not a separate tool beside it. The same approach can be adapted to a new operation, and Athreix aims to show a first working proof within 48 hours.
Questions about EdgeOpt
What is EdgeOpt?
EdgeOpt is a model optimisation for devices by Athreix in its Computer vision & applied AI division. Prepare models for constrained devices and compare the result with the baseline.
How does EdgeOpt work?
EdgeOpt works in 3 steps: review the model, model compression + device conversion and compare + export.
Can Athreix build something like EdgeOpt for my business?
Yes. Athreix adapts systems like EdgeOpt to each operation, starting from the workflow, the data and the result you need, with a first working proof aimed for within 48 hours. Contact tech@athreix.com.
Every Athreix system ships with a human in control, defined fallback behaviour and a record of what it decided.
