Building things that make complex systems understandable
AI-accelerated observability and performance engineering for organizations navigating complexity. Based in Sweden.
"The liminal space is where transformation happens — the threshold between what was and what will be."The philosophy behind Proliminal
Services
AI Integration
Helping organizations adopt AI meaningfully — from strategy to implementation. Agent architectures, LLM integration, and AI-first development frameworks.
Performance Engineering
Making complex system dynamics visible, comparable, and optimizable. Bespoke load testing, capacity planning, SLA/SLO analysis, and architectural risk assessment — from early design through production.
Systems & Observability
Making complex systems understandable through visualization, flow modeling, and developer tooling. From architecture to the user interface.
UX & Usability
Interfaces that serve people. Bringing empathy and usability thinking to technical products — because good systems deserve good interfaces.
Architecture
Distributed systems and event-driven architectures. Building foundations that scale with clarity.
Transformation
Guiding teams and systems through the liminal space — from legacy to modern, from manual to AI-native, from opaque to observable, leading by example.
How I Work
Proliminal takes its name from liminality — the transformative space between what was and what will be. Whether working early in architecture to identify risks, or close to production to find real bottlenecks, every engagement follows this arc:
Understand
Instrument, measure, and model. Find where assumptions about the system diverge from actual behavior. Identify the real constraints and the real opportunities hiding in the flows.
Transform
The liminal phase. Build bespoke tooling, test hypotheses with real data, iterate fast. AI-accelerated delivery from design through validation. Embrace productive chaos.
Emerge
Arrive at clarity. Dashboards leadership can read, flow models teams can reason about, and before-and-after evidence that justifies the investment. Knowledge transferred. Autonomy restored.
The tools and methods in the Lab aren't theoretical — they inform how I approach client work. Provenance thinking shapes audit trails. Flow modeling informs performance analysis. The same architectural rigor applies to consulting engagements. Lab →
Selected Work
Biometria
Six years of performance engineering on a mission-critical platform handling ~1M daily transactions. Built a bespoke observability and load testing platform, introduced queue-centric flow analysis, and reduced critical document lead times from ~1 hour to under 1 minute.
Liminara
Open-source provenance runtime making computations reproducible, auditable, and verifiable. Cryptographically sealed evidence trails for EU regulatory compliance.
FlowTime
Flow analytics and simulation platform that makes operational systems understandable — replay what happened, run what-if scenarios, and identify bottlenecks before changes hit the real world.
A2UI Protocol
Agent-to-UI protocol implementations — agents send declarative JSON component trees; the client renders approved widgets.
Agent Lens
VS Code extension that makes AI coding agents observable. Works with GitHub Copilot and Claude Code.
AI-First Framework
Portable AI-assisted development framework with role-based agents, reusable workflow skills, and guardrails.
Daklapack Europe
Mission-critical production system for medical self-test kits. When a legal freeze killed all Azure assets, rebuilt the entire system on-prem in days — zero downtime, zero tolerance for errors.
About
Peter Bruinsma is a consultant and technologist with over 30 years of experience across the USA, the Netherlands, and Sweden. He started his career at Microsoft in Redmond, later returning to Microsoft in the Netherlands as an application development consultant. In 2000 he founded CraitonTesting AB, an independent test house specializing in hardware and system certification for international manufacturers, which he ran for 13 years.
He spent six years with Biometria — Sweden's central measurement organization for the forestry industry — doing performance engineering and systems work on mission-critical platforms handling over a million transactions per day. Most recently, he designed and launched a bespoke near-real-time observability platform serving SRE, developer, architect, and executive audiences.
Today, his focus has shifted heavily toward AI — not just as a technology to integrate, but as a fundamental shift in how we build, observe, and interact with software systems. His own delivery is now fully AI-native, using AI end-to-end from planning through coding, testing, and deployment. His open source work reflects this: tools for AI agent observability, agent-to-UI protocols, and frameworks for AI-first development.
He believes that the best technology is technology you can understand. This conviction drives everything from his approach to systems architecture to his deep interest in UI/UX and usability. Good interfaces are acts of empathy.
The instinct started early. During his Microsoft internship in 1991, he designed the Wizard UI pattern, on paper, before it existed in any product. Printing and charting in Excel were painful multi-step processes. The Wizard made them simple by guiding decisions one step at a time. It went on to become one of the most widely adopted UI patterns in software history. It started with noticing where people were struggling.
About the name ProLiminal →Ask Proliminal
AI-Powered Q&A
Focus Areas
- AI & LLM Integration
- Performance Engineering
- Observability
- Agent Architectures
- UI/UX & Usability
Technologies
- C# / .NET / Blazor
- Elixir / OTP
- Microsoft Azure
Tools
- Claude Code
- GitHub Copilot
- OpenAI Codex
Location
Sweden
Previously: USA · NL
Background
Ex-Microsoft · Founded CraitonTesting AB (2000–2013)
Company
Proliminal AB
Contact
Looking for help with difficult optimization problems where standard tooling isn't enough, from someone who brings both deep technical expertise and genuine care for how systems serve people? Let's talk.