Miniml is headquartered in Edinburgh, and two of our founders are professors at the University of Edinburgh. For Scottish organisations that means the people who publish the research sit a short walk from your office, and they stay on the project until the system is running.
Our office is on George Street, in the middle of Edinburgh’s financial district. Discovery workshops, integration sessions and go-live happen in person as a matter of course, and a senior consultant can be at your site the same week.
Edinburgh’s strengths shape our work here: one of Europe’s largest asset-management and banking centres, a fintech community that has grown up alongside it, a health and life-sciences cluster at the BioQuarter, and a public sector that runs much of Scotland from the city.
Both of our published case studies are Scottish. That is not a coincidence; it is where we live and work.
Speech recognition tuned to clinical language, notes in the board’s templates, SNOMED CT coding aligned to its standards. Four milestones delivered on schedule, no real patient data used.
Read the case study →A Scottish brokerage’s inbox, documents and case system handled by our agents in production, with advisers approving every consequential action.
Read the case study →Our CTO holds a chair in natural language processing and our CSO in machine learning, both at the University of Edinburgh. Our CSO is also a Turing Fellow. They lead client work, not just research.
Meet the team →Edinburgh is a compact city with an unusual concentration of regulated, data-rich organisations. These are the sectors where we do most of our Scottish work.
Edinburgh’s investment houses and banks produce and consume an extraordinary volume of documents: research, fund reports, policy, client correspondence. We build retrieval systems that let analysts question that corpus with citations, and drafting assistants that produce first versions of recurring reports for a person to finish, inside the firm’s own environment.
For the lenders, brokers and payment firms around the city, the operational load is in communications, documents and case handling. We have built exactly this in production: agents that classify incoming client communication, extract structured data from documents and prepare approved updates in the case system.
Working with an NHS health board in Scotland, we built a medical scribe that turns consultations into structured, coded notes, designed to run on NHS infrastructure and developed without real patient data. The BioQuarter and Scotland’s health-innovation bodies are part of our working world, not a distant sector.
Government agencies and the universities headquartered in Edinburgh handle casework and public correspondence at volume, under procurement and transparency rules. We build assistants that draft, route and summarise with a full audit trail, and knowledge systems that make policy and precedent searchable for staff.
Every capability has its own specialist page. In Edinburgh the work skews towards regulated document handling, clinical AI and knowledge systems for large institutions.
Speech recognition tuned to clinical language, models adapted to a firm’s own terminology, and evaluation built from your real cases. Our Scottish NHS work is the reference point: a prototype delivered on schedule across four milestones.
Permissioned retrieval over research libraries, policy archives and case files, with every answer cited. Built for institutions where access control is as important as accuracy.
Agents that process communications and documents and prepare actions in your core system, with a person approving anything consequential. Running in production for a Scottish financial-services client.
For NHS boards, public bodies and financial firms whose data cannot leave a controlled environment: open-weight models adapted to your domain and run inside your own boundary.
An evidence-led roadmap sequenced by value and risk, written for executive committees that need to fund one thing this year rather than experiment with ten.
Most of Edinburgh’s largest organisations cannot send data to an external AI service: asset managers under FCA rules, NHS boards under information-governance standards, public bodies under procurement and transparency law. That constraint used to mean waiting. It no longer does.
We adapt open-weight models to your domain and run them where your rules say they must run. In 2026 those models are within a few percentage points of the best hosted ones, the hardware to run them fits in a rack or a private cloud tenancy, and the audit trail is yours. Our NHS scribe work was built exactly this way, for NHS infrastructure, without any patient data leaving the board.
Model, retrieval and evaluation on your own servers, whether that is an NHS data centre, a bank’s estate or a university facility. Nothing leaves the building.
A dedicated environment in a UK region under your keys and access control, for organisations that need residency and auditability without owning hardware.
Inside the tenancy your IT team already governs, inheriting its identity, monitoring and change control rather than adding a new perimeter.
A discovery session at your office or ours on George Street, with the people who own the workflow. We leave with a scored shortlist and a named business owner.
A working prototype on your documents and systems within weeks, measured against a test set from real cases, with our engineers available on site as often as it helps.
Connection to your systems of record with approvals, audit trail and monitoring, designed around your governance, whether that is FCA, NHS information governance or public-sector procurement.
Go-live with your team trained to run and extend the system. Being in the same city, we stay close through the first months of operation.
Each is led by a senior consultant and ends with a written result you can act
