Service / 12
Research
We take on applied research mandates for organisations that need evidence before they commit: model evaluations, feasibility studies, benchmarks, and technical due diligence. Findings are reproducible and written to be acted on.
When you need this
The situations we are called into.
You must choose between models, vendors, or architectures — and public benchmarks do not answer your question.
An acquisition or investment requires technical due diligence on an AI product.
A regulator, standards body, or court needs the evidence behind an AI claim.
Approach
How the engagement runs.
Question formulation
The decision the research must inform, stated precisely — so the study answers it and nothing else.
Method design
Datasets, metrics, baselines, and controls specified and agreed before any experiment runs.
Experimentation
Runs executed on versioned code with logged configurations, so every number can be regenerated.
Analysis
Results with uncertainty stated, limitations named, and alternative explanations addressed.
Reporting
A technical report and an executive summary, presented and defended to your stakeholders.
Deliverables
What you hold at the end.
Study design document
Reproducible experiment code
Technical report
Executive summary
Findings presentation
Related research
The evidence behind this practice.
Technical report · 2026
Evaluating large language models for regulated enterprise workflows
A practical evaluation framework for LLM systems in banking, insurance, and government contexts — metrics, test harnesses, and a taxonomy of failure modes.
In preparation
Benchmark · 2026
Private LLM inference on Apple-silicon clusters
Throughput, latency, and cost characteristics of quantised open-weight models served on commodity Apple-silicon hardware, measured against cloud baselines.
In preparation
White paper · 2026
Multi-agent systems in production: reliability patterns
Failure modes observed when orchestrating multiple LLM agents on business-critical tasks, and the containment patterns that keep them recoverable.
In preparation
Contact
Talk to our engineers.
Describe the situation you are in. The person who replies is the person who would do the work.