Robin Miklinski Independent AI product and systems engineer · London
I work with founders, product teams and established businesses. I make the important technical decisions, work in the code and bring in specialists when the project needs them.
Previously
Reward Gateway ASOS Perkbox Huddle BAE Systems Detica
You have an AI product or prototype and need the engineering required to release it. I work on the architecture, data and integrations, then define how the product will be evaluated, secured and operated.
I can join an existing product team, lead a defined part of the work or take technical responsibility for the project through release.
An important process relies on manual hand-offs, disconnected tools or software that no longer fits the work. I start by understanding the process, the information it depends on and the points where time or accuracy is lost.
The answer may be an integration, a focused software tool, an AI-assisted step or no AI at all. Once the change is clear, I design it and contribute to the implementation with the client team.
I decide how the product should be structured, how data moves through it and how it connects to the software and services around it.
I turn product requirements and technical risks into concrete checks. For AI products, that includes evaluating outputs, testing misuse and failure cases, and deciding what evidence is needed before a release. I record what the testing did not establish alongside what it did.
I make the important technical decisions and work directly in the code. I work with the engineers already responsible for the product and leave them able to continue without relying on me. When a project needs deeper expertise in a particular discipline, I bring in the right specialist and keep the responsibilities and technical direction clear.
A local-first release-planning application that runs in the browser without accounts or a backend.
Source →A background audio service for metadata cleaning and BPM detection that leaves its source files unchanged.
Source →An OWASP ZAP client for automated security scanning in continuous integration.
Source →AI product evaluation. I created the evaluation and adversarial security testing for the company’s first customer-facing AI product. It covered answer quality, prompt injection, tenant isolation, personal data and retrieval scope.
Incident analysis. I built a Python pipeline across the company’s incident, documentation and monitoring systems. It analysed 451 production incidents over ninety days and supported a £227,000 annual business case for preventable failure.
Delivery infrastructure. I designed the systems coordinating roughly one thousand end-to-end tests across pull-request, release and nightly runs.
I also built and led the quality engineering function that operated these systems, growing it from zero to fifteen engineers across London and Sofia.
I took over a Dart, Flutter and Firebase product with little documentation, integrated a planning data API, restored a failed council dataset and rebuilt its scoring system. I then re-scored more than 24,000 planning applications and onboarded the external development team that continued the work.
I was Head of Quality Engineering at Reward Gateway. The role combined technical leadership with the delivery, evaluation and incident-analysis systems described above.
My experience spans defence and intelligence data systems, retail at peak scale, collaboration software and global SaaS used by millions. I work primarily in Python, JavaScript, TypeScript and C#, and I hold a BSc (Hons) in Internet Engineering from the University of Exeter.
Outside engineering, I DJ and produce electronic music as Eidetic. I have been a resident at EGG London and hosted Late Night Relics on AAJA Radio.
Tell me what you are building, how the work happens today and where the current system stops holding up. I’ll respond with the questions, risks and next step I think matter.
Prefer email? robin.miklinski@gmail.com