How we got here
Secure AI Dynamics started when two machine-learning engineers left a large consultancy because they were tired of writing proof-of-concept notebooks that never made it into production. The idea was simple: only take on projects where the model will actually run in a live system and generate measurable value.
The first client was a Belfast-based courier company struggling with late deliveries. We built a route-planning model trained on three years of GPS logs and weather data. Within four months their on-time rate went from 81% to 93%. That project paid for our first office on Pound Lane, where we still work today.
Six years later the team has grown to 18 people. We have delivered 127 models across retail, healthcare, logistics, insurance and legal services. The company remains privately held and debt-free, which means we pick projects based on fit rather than revenue pressure.
What we believe
Models belong in production
A model sitting in a Jupyter notebook helps nobody. Every engagement we take on ends with a deployed, monitored system that your team can operate without us. We write runbooks, not just research papers.
Data stays under your control
We work inside your environment whenever possible. When data must leave your servers, we anonymise it first and delete our copies within 30 days of project completion. No exceptions.
Fixed prices, no surprises
After a two-week discovery phase we quote a fixed price for the entire build. If scope changes mid-project, we re-quote before continuing. You will never receive an invoice you did not approve in advance.
Honest about limitations
If your data cannot support the outcome you want, we say so during discovery rather than billing for months of dead-end experiments. Roughly one in five discovery audits ends with a recommendation to collect better data before starting a build.
The team
Eighteen people across engineering, data science and project management. Here are the four who lead client engagements.
Dr Fiona Gallagher
Co-founder and technical director. PhD in statistical learning from Queen's University Belfast. Leads our NLP and document-processing projects.
Marcus Chen
Co-founder and head of engineering. Former infrastructure lead at a FTSE 250 logistics firm. Responsible for deployment pipelines and uptime.
Amara Osei
Senior data scientist. Specialises in computer vision for manufacturing and retail. Previously built defect-detection systems at a semiconductor fab.
David Hargreaves
Client engagement manager. Runs discovery audits and translates business requirements into technical specifications. 15 years in IT consulting before joining us.
Our timeline
2019 — Founded
Fiona and Marcus registered the company and signed the courier-route project that became our first production deployment.
2020 — First healthcare client
Built a clinical-note summarisation tool for a GP federation in Belfast, processing around 800 consultation notes per day.
2021 — Team reaches 10
Hired dedicated MLOps engineers so every model ships with monitoring, alerting and automated retraining triggers.
2022 — Retail expansion
Deployed demand-forecasting models for a 140-store grocery chain, reducing overstock waste by 17% in the first year.
2023 — Security audit service launched
Started offering adversarial testing and bias audits for organisations already running their own ML models.
2024 — 100th model deployed
Crossed the milestone of 100 models running in production across six industries. Team grew to 18.
Get in touch
Visit us, call or send an email. We are happy to have an initial conversation at no cost.
United Kingdom, Northern Ireland, Rodriguezbury, CD69 6VO, 247 Pound Lane