Services
Four disciplines, one standard of engineering.
Each engagement is led by the engineers who build it. Most projects draw on more than one discipline; the written scope says exactly which.
Machine learning
Models for forecasting, scoring, ranking and optimisation. We agree the business metric first, then build, test and deploy against it.
Typical problems
Demand and inventory forecasting
Lead, risk and churn scoring
Scheduling, priority and assignment models
Simulation and digital twins
What you receive
A model evaluated against your business metric
A deployed API or scheduled pipeline
A monitoring and retraining plan
WEEKLY DEMAND · STORE 14
−12%
stock-outs this quarter
Fig. 1 — A demand dashboard from a retail forecasting build.
Computer vision
Systems that see: detection, segmentation, counting, measurement and inspection — from site surveys to shop floors.
Typical problems
Object detection and counting
Mapping and measurement from imagery
Visual quality inspection
Event detection on live video
What you receive
A labelled dataset and labelling guide
A model with a field accuracy report
Edge or cloud deployment
Site survey, auto-traced
6 boundaries · 0.4 m accuracy
Fig. 2 — A site survey traced and measured by a vision model.
AI agents & automation
Agents that read, decide and act inside your tools, with defined permissions, full audit logs and human review where the stakes are high.
Typical problems
Document intake and data entry
Support and operations triage
Research, reporting and summaries
Workflow automation across systems
What you receive
An agent with defined tools and permissions
An audit log and human review steps
A runbook for your team
Morning. I read 18 new invoices and matched 16 to purchase orders.
2 don’t match. They need your eye before payment.
INV-2291 · £4,120
PO amount differs by £380
Fig. 3 — An intake agent asking a person to approve two invoices.
Software & infrastructure
The engineering around the model: backends, data pipelines, internal tools and the cloud setup that keeps it all running.
Typical problems
Backend APIs and data pipelines
Dashboards and internal tools
Cloud setup, CI/CD and monitoring
Integrating models into existing products
What you receive
Documented code in your repository
Infrastructure as code
A handover session with your team
Fig. 4 — The architecture sheet handed over with every build.
What we build with.
We choose tools for the job and for your team’s ability to maintain them after we leave.
Languages
Python, TypeScript, C++, SQL
Machine learning
PyTorch, scikit-learn, XGBoost, OpenCV
Agents & language models
Hosted and open-weight models, tool calling, retrieval
Data & backend
FastAPI, PostgreSQL, Redis, message queues
Cloud & delivery
AWS, Google Cloud, Azure, Docker, Terraform, GitHub Actions