Low compute
Models that run on microcontrollers, edge CPUs, and embedded accelerators with tight memory and processing limits.
ML models that run within your constraints.
We specialize in small, efficient models for environments where generic cloud ML can't reach — low compute, low power, offline, or privacy-sensitive. Fixed deliverables, fixed timeline. If we don't hit the spec, you don't pay.
How it works
-
Discovery
We look at your product, your data, and public datasets to find where ML can create real value.
-
Proposal
A scoped project with concrete deliverables, a fixed price, and a fixed timeline.
-
Delivery
You only pay if what we ship matches the spec we agreed on.
What we specialize in
Low power
Efficient enough for battery-powered and energy-harvesting devices where every milliwatt counts.
No network
Fully offline inference for remote sites, mobile assets, and environments without reliable connectivity.
Privacy & security
Data stays on the device or inside your network. No round trips to the cloud, no third-party exposure.
Typical projects
- Defect detection on inspection drones
- Anomaly detection on industrial sensors
- Object recognition for underwater robotics
- Predictive monitoring on grid equipment in the field
Who this is for
Teams building products where ML needs to work within real constraints — hardware limits, latency requirements, data privacy, or operating conditions that rule out a generic API. We work with OEMs, robotics companies, device manufacturers, and asset operators.
Pricing
£30,000 – £200,000
The range most projects fall into, agreed upfront. A focused proof of concept sits at the lower end; a production-ready deployment sits at the upper end.
Get in touch
hello@ontosyn.comTell us what you're working on. We'll come back within a week with a view on whether there's a project here.