01 / 06 · Showcase
TeamFlow
The DevOps harness for agentic software delivery.
Observability, automation and agentic SDLC management: every agent workstream on one board, stuck agents found and fixed, and code that never leaves your organisation.
AI Product Rescue
I make agentic AI shippable.
02 / 06 · Showcase
Provely
The defensive layer for AI agents.
A workflow verifier for AI agents that take side-effecting actions. A separate verifier decides whether reality satisfies the promise, and issues a signed receipt.
03 / 06 · Research
Heart
A causal 4D beating heart with a live 12-lead ECG.
Real CT-derived four-chamber anatomy inside an MRI-segmented body, driven by a physiological activation model and rendered live in the browser with a synchronized, model-derived 12-lead ECG.
04 / 06 · Research
Doc
Probabilistic causal diagnosis.
A probabilistic causal-graph physician: exact noisy-OR inference over a causal knowledge graph, with decision-theoretic stopping that picks the most informative next question.
05 / 06 · Product Beta
Grow
Your product deserves users.
A developer growth OS in beta: Grow researches your product, rebuilds and optimizes your site, generates and distributes content, studies competitors and measures real outcomes, 24×7.
06 / 06 · Product
MechKit
The $1 mechanical solve.
Send a bounded component-and-claim model; get back a reproducible layout that passes the declared checks, or the exact constraints still in conflict. An LLM geometry guess is never a verified result.
From the lab
Featured field note
The Manager as Compiler
WorkSpec: the missing control layer for AI work
AI does not just change how work is produced. It changes how work must be specified, reviewed, constrained, and accepted. The problem is no longer output. The problem is control.
Read the article
Recent clients
CTO and Chief Architecture roles
[Steve was] asked to rectify the bugs in the code, move it from a test environment to production, document the code and accounts used, and recommend the development tools that would suit the AI development that was taking place. Whilst the first two weeks was impressive, the achievement of these stretch goals was even more impressive, as we were left with a fully documented and running environment, which made it to market in tight deadlines.
— Chief Technology Officer, AI-security company
Named client references available on request.