Every project we take starts with the same two questions: what do you measure, and what decision hangs on it? Between those two points there is always a model - and the difference between a model that gets used and one that gets quietly turned off is everything this page is about: features that carry your domain's structure, predictions with calibrated confidence, explanations an expert can interrogate, and a deployment that learns from its mistakes.
The way we get there is a house method - foundation models for features, fast models to ship, conformal calibration on top - described plainly on how we build.
Computer vision
Microscopy, inspection lines, field cameras, lab imagery - most scientific and industrial data that matters is a picture of something. We build detection, classification and measurement systems on modern vision backbones, including self-supervised models that learn your domain's structure from unlabelled archives before the first annotation is made. Small objects, rare defects, imbalanced classes - the unglamorous cases that decide whether vision works in production are the ones we specialise in. And because the backbone yields features, not just answers, every system comes with attention maps and attributions that show what the model looked at - so an expert can tell a real detection from a lucky one.
Language & agents
The newest member of the toolbox, held to the oldest standard: measured behaviour. Assistants grounded in your documents, structured extraction from decades of PDFs, generated reports an expert signs off on, and agents that act with scoped permissions - all built evaluation-first, so "how often is it wrong" has a number before anything ships.
Prediction & forecasting
Quality from process parameters, properties from compositions, demand from history, anomalies from sensor streams. On tabular and time-series data we ship small models on engineered and learned features - trained in minutes, explainable by construction, and wrapped in conformal intervals with a guaranteed error rate. When the model says 90%, it is right nine times out of ten; that is a property we deliver and verify, not a slide.
How we can help
Vision systems
Detection, measurement and anomaly finding on scientific and industrial imagery, with explanations attached.
LLMs & agents
Retrieval, report generation and tool-using agents - evaluation harness first, demo second.
AI Trust & Security
Red-teaming and audits for the LLM systems you already shipped - including ones we did not build.
Model validation & deployment
Is the model you already rely on calibrated, honestly evaluated, and reading the physics rather than an artefact?
The method
Foundation models, domain adaptation, fast models, conformal calibration - why our systems are quick to retrain, explainable and honest about uncertainty.
Deployment
Pipelines, Kubernetes and the active-learning loop that makes a deployed model improve instead of decay.
Workshops
Hands-on sessions for your team, tailored to your applications and - on request - your own datasets.
Looking for the earlier service pages - statistics, classical machine learning, deep learning? They are still there, at the same URLs as always, marked as archived. They are no longer in the menu, because they describe what we sold in 2017 rather than what we do now.