Hi, I'm Peter McCann Strain.

The short version

I take hard scientific and technical problems and turn them into products people can actually rely on. The work ranges across AI ventures, enterprise systems, legal infrastructure, medical AI, satellite vision, and the odd private R&D thread that teaches me something useful before it becomes public.

The throughline is not “AI everywhere.” It is careful engineering, research-grade judgement, and the discipline to know what is worth shipping. I like systems that can be inspected, tested, challenged, and used by real people without pretending the messy parts of the world do not exist.

Another way to put it: Oxford gave me the research habit, founder work taught me that shipping changes the question, and enterprise AI made control, auditability, and cost impossible to treat as afterthoughts. I keep coming back to the same thing: build the useful thing, make its limits visible, and leave people enough to trust or challenge it.

What I am building now

Right now I'm building in two directions: Praviar and Conversico. Praviar is pharmaceutical patent-intelligence support: AI helping with freedom-to-operate research while fixed checks and expert review keep the conclusions bounded. Conversico is real-time voice operations for dental practices: missed-call recovery, booking support, privacy-conscious workflows, operator visibility, and graceful failure when the system should not pretend.

I treat both as founder-building work: product taste, systems engineering, launch discipline, and restraint in domains where overclaiming would be worse than useless.

The work behind it

At SThree, I was one of two founding AI engineers and architected Aurum as employer work: enterprise AI for recruitment built from teams of AI agents working together. Aurum coordinates those agents, generates realistic test candidates, scores its own output and learns to score better over time, keeps every step reviewable, checks for bias, stress-tests itself against deliberate attacks, and stays controllable in production.

Through Datamise, which I co-founded in 2023, a two-person team built owned equal-pay claims infrastructure for GMB Union's campaigns. We then extended the work with SettleMise: a reviewable valuation engine that supports human legal and accounting review; it does not decide liability.

Before SThree, at Astroscale, I built computer-vision systems that work out a satellite's exact position and orientation in space, and designed AstroGAN, which turns cheap simulated images into ones that look convincingly like real satellite photos — closing most of the realism gap so the systems train on better data. My Oxford DPhil built AI that reads full digitised microscope slides of mouse placental tissue: spotting individual cells, sorting them into eight types with 89% accuracy, and mapping out tissue regions.

The quieter threads

Personal Assistant AI is the private product R&D thread behind an executive-assistant direction: a daily-use assistant for capture, durable memory, proactive review, action, and quality monitoring. It is deliberately presented as private single-user dogfooding, not customer traction. The deep-research benchmark and Aurum evaluator work add the rigour layer: checking that the automated scorers can be trusted, verifying every cited source, testing fairly for genuine equivalence, guarding against data leaks, tracking each claim, and reporting the results that did not work as well as the ones that did.

Skills Taxonomy is supporting production-economics craft rather than headline material: I rebuilt SThree's system for tagging the skills on a CV so it no longer calls a large language model every time it runs — instead it looks up close matches, applies fixed rules, and uses a tiny, fast classifier. The point is judgement under cost and speed pressure: knowing when a simpler, inspectable system is the stronger production move.

On the public-sector side, I was selected for the CivTech Exploration phase as a sole individual applicant. In three weeks I produced a comprehensive accelerator plan for a human-in-the-loop AI system covering Scotland's eight statutory impact-assessment workflows. The plan did not progress to the next phase, so I present it as exploration-stage product strategy rather than a delivered government system.

Press & media

Short bio, for reuse: Dr Peter McCann Strain is a founder and AI builder who takes genuinely hard scientific and technical problems and turns them into products people can rely on. He works across his own AI ventures, agentic systems, legal-tech, and computer vision. He holds an Oxford DPhil in AI for medical imaging, was a founding AI engineer at a FTSE 250 staffing firm where he architected an enterprise agentic-AI platform, and co-founded the consultancy Datamise.

A high-resolution portrait and a full CV are available on request. For interviews or written commentary, email is the fastest route.

Get in touch

If there is a thread here that connects with something you are building, researching, or trying to understand, I would be glad to hear from you. That might be a senior or founding role, a build partnership, a collaboration, or simply a hard problem worth thinking through properly.

peter.mccann.strain@gmail.com · LinkedIn

Looking for the short version? Head back to the homepage.