Forecasting, vision, ranking and anomaly detection that ships.
What this actually involves
Classical machine learning still wins on a great many problems, more cheaply and more predictably than a language model. We pick the smallest technique that clears the bar.
Models are delivered with the boring apparatus that keeps them honest: drift monitoring, retraining triggers, a champion-challenger process and a documented rollback.
Computer vision
Detection, segmentation and OCR, including edge deployment on commodity hardware.
Forecasting & optimisation
Demand, capacity and pricing models with quantified uncertainty.
Ranking & personalisation
Retrieval and ordering that improve a metric you agreed in advance.
Anomaly detection
Fraud, defect and fault detection tuned to your cost of a false positive.
What lands in your repository
Every engagement ends with artefacts your team owns — not a slide deck describing artefacts your team could have owned.
- Trained model and inference service
- Offline and online evaluation results
- Drift monitors and retraining triggers
- Model card and limitations statement
A short conversation with an engineer, not a sales qualification call. If we're the wrong people for it, we'll say so and point you somewhere better.