AI Developer
You’d build the models that decide what’s worth looking at in an hour of drone footage — and then make them run on the device, not a server somewhere.
Real footage, not a clean dataset
Our AI Vision App runs object and anomaly detection on payload footage — live during a mission or on something recorded earlier — and flags what actually needs a human look. The training data is real field footage: monsoon light, dust, motion blur, all of it. Getting a model that works on that, and then getting it small and fast enough to run on-device instead of a cloud GPU, is the actual job.
We made a deliberate call not to build a cloud backend for this. If that constraint sounds annoying rather than interesting, this probably isn’t your role.
What this actually involves
- Training and iterating on detection/anomaly models using real mission footage
- Quantising and optimising models to run on-device with acceptable latency
- Working with the integration and embedded teams on what the hardware can actually support
- Reviewing false positives and false negatives from live trials, then going back and fixing the model
- Keeping the whole pipeline — training to deployment — something one person can actually reason about
What we’re looking for
Less about a specific degree, more about having actually shipped something.
Show us something you’ve trained
A GitHub link, a Kaggle notebook, a side project — anything where you took a model from idea to actually running is worth more to us than a polished resume.
This isn’t the only seat open
Four other roles are open right now, from hardware integration to embedded firmware.