Research
Research that has to survive the field.
Three programmes, one constraint: the system has to work where data is thin, the network is slow and the decision cannot wait.
Three programmes in progress. No paper is public yet.
Programmes
What the lab is working on.
Each programme is tied to a product and to a deployment, not to a demo.
- 01In progress
Agricultural Intelligence
AI-powered systems for farming, from soil analysis to yield prediction.
- 02In progress
Predictive Systems
Forecasting water stress, diseases, yields and supply-chain risks.
- 03In progress
Language & AI
Models and systems that understand context, in African languages and beyond.
Method
Field first, then the model.
A model that cannot be checked against the field is a guess with better formatting.
- 01
Start from the record
We look at what operators already write down, and where the record breaks.
- 02
Model the constraint
Water, distance, language, connectivity: the constraint is part of the design.
- 03
Test in place
Evaluation uses the deployment's data and conditions, not only a clean validation set.
- 04
Publish the limits
A result says where it stops being true. That is what makes it usable.
Notes
Written as the work moves.
Field notes and product notes from the teams. No calendar.
Intelligence for the real world
Why Gytev is not building a search engine, but the layer that connects digital intelligence to reality.
Sovereign AI
Why we build it here.
Our data and compute stay in West Africa. Read why that is the default.