Private AI · Lausanne / Geneva
AI for sensitive work.Keep the data. Choose the model.
I build practical AI systems around documents and workflows where confidentiality matters. Run models locally when information should stay in-house; use hosted models when that is the right trade-off.
The principle
The model is replaceable. Your data and workflow are not.
- Deliberate boundariesChoose where sensitive information is processed and stored.
- Useful before cleverAutomate a real task instead of adding a chatbot for its own sake.
- Visible systemsSmall, inspectable software with clear boundaries and human review where needed.
What I build
Three practical layers.
The same approach scales from a local script to an internal application.
Private AI
Local LLMs, retrieval, document processing and controlled access for information that needs a deliberate boundary around where it is processed and stored.
Document intelligence
Turn the PDFs, DOCX files, transcripts and other material already in your organisation into searchable, summarised and usable information.
Workflow software
Small applications and automations that connect the AI to the work around it — with humans retaining approval where it matters.
Try it now
See the mechanics.
A simple example of the kind of document-handling problem I build around.
Marked up
Explore the token counter, document splitter and other labs →
Evidence
Things that already run.
DocSumm
Local summarisation for PDF, DOCX, HTML, EPUB and meeting audio. Designed to sit beside a local model.
CV Analyzer
AI-assisted CV and job-description analysis that surfaces ATS gaps and helps turn structured career data into targeted applications.
The Metric
Structured career data and software for targeted applications, interviews and career decisions.
For organisations handling sensitive information
Start with the workflow, not the model.
Bring the sensitive document, repetitive process or awkward hand-off. We work out what should stay local, what can be automated, and where a human needs to remain in control.
Who is Rametric?
One experienced builder, not a fictional agency.
I am Michael Coughlan. I have worked across product, programme delivery and software, and now build focused AI systems around real documents and workflows.