Daniel Patyński
Senior engineering, without the agency.
Independent software and AI delivery consultant with eleven years shipping production systems for UK government, retail, automotive and finance. Whether it's an AI demo that has to survive real users or a system that's slow, broken or late, I get it into production and leave your team able to run it.
Through consultancy engagements and in-house roles, 2015 to today.
Pick the closest match. If your problem isn't here, ask anyway: most of my best work started as "we're not sure who to call about this".
LLM features that hold up with real users: structured outputs, retrieval and citation grounding, evals, guardrails and cost control. In Contractalyst I built contract analysis this way. Any clause the model can't quote from the source document is rejected before a user sees it.
Good fit if you have a promising AI demo and need it to be reliable, auditable and affordable.
Spec-driven, agent-assisted delivery: repository rules, CI gates and review habits that let engineers use Claude Code and similar tools without lowering the bar. Delivered as a hands-on workshop, or embedded with your team for a few weeks.
Good fit if your engineers use AI tools but the results are inconsistent, or leadership wants a clear AI policy.
Timeouts, N+1 queries, memory leaks, flaky pipelines, code nobody wants to touch. I find the actual cause, fix it, and leave tests behind so it stays fixed.
Good fit if something is slow, failing or late and nobody can say why.
From idea to launched product: architecture, backend, frontend, auth, payments and CI/CD. Java or Kotlin with Spring Boot, TypeScript and React, PostgreSQL, AWS. Built so your next hire can pick it up.
Good fit if you need an MVP or internal tool built properly the first time.
Identity and access management (ForgeRock, OAuth and OIDC, JWT, role-based access), cloud migrations under strict security requirements, and GDPR work. Years of experience on UK government services and security-sensitive platforms.
Good fit if auth, compliance or a security review is blocking your roadmap.
Numbers from production, measured on real systems I worked on. The full bar is before; the bright part is what's left.
Finance automation platform for enterprise clients
150 s 11 s
It used to hit the gateway's 150 s limit and fail. Rebuilt as batched streaming, it now finishes in 11 s.
Identity platform used by a dozen teams, HM Courts & Tribunals Service
30 s 2 s
Reverse-engineered the slow path in a customised IAM solution and removed it.
Approvals inbox taking 50 s to load
1,001 2
Traced the 50 s load to an N+1 query pattern. Fixing the root cause took it to 2 queries.
On a security-sensitive identity programme, the client made its continued engagement with the consultancy contingent on keeping me on the team.
A multi-tenant SaaS for construction contract lifecycle management, designed, built and launched on my own. It's live and entering its pilot. It shows what I bring to a new build: the whole stack, and the boring parts done properly.
No sales team, no account managers. Here's how an engagement goes.
Free. You describe the problem and I tell you honestly whether I'm the right person for it.
I look at the code, the system and the constraints, then write down what's wrong and what I'd do about it. Fixed scope, so you know the cost up front.
Hands-on. I write the code, embedded in your team or against an agreed scope, with progress you can see every week.
Tests, documentation and a team that understands what was built. You shouldn't need me afterwards.
I'm Daniel Patyński, a software engineer who has spent more than eleven years building and fixing backend-heavy systems. I started on document publishing platforms for Ford, Jaguar Land Rover, JCB and Harley-Davidson. I then spent five years as a consultant at Ensono (formerly Amido) on HM Courts & Tribunals Service, Screwfix and UK pension services, became a staff engineer at Leidos leading an identity platform's cloud migration, and most recently have worked on enterprise finance automation.
Along the way I've learned that most software problems aren't about technology. They're about nobody owning the hard part. I own the hard part.
First Class BSc in Computer Science from Kingston University. AWS Certified. Outside work I captain a table tennis A-team and spend too much time on motorcycles, Linux and 3D printing.
Running an event or a team day? I've put together a talk and workshop called Using AI to deliver code: why it's not cheating, it's engineering. Ask about it below.
A few lines is enough. I reply within one working day.