AI Engineering Assessment
A written assessment of an AI workflow, prototype or product idea: where it adds value, what can fail, and what needs to exist around it before people depend on it.
Technology that creates value, not overhead.
Maybe you’re working out where AI genuinely helps, or a promising prototype now has to survive real users. Maybe your team needs clear technical direction through the shift. That’s the work I do with startups, scale-ups and SMEs.
Behind it: 16+ years building distributed systems, six of them leading engineering on the core payments platform of one of the UK’s largest fintechs, and the last two building with LLMs and agents.
How I Can Help
Not every engagement starts with a project plan. Sometimes it starts with a question, a challenge, or a decision that needs careful thought.
Engagements range from a focused second opinion to building and shipping a production system end to end.
A written assessment of an AI workflow, prototype or product idea: where it adds value, what can fail, and what needs to exist around it before people depend on it.
A production-grade system taken from scope to handover: designed, built, integrated and ready to run.
A structured working session that turns an unclear technical problem into options the team can evaluate, with trade-offs made explicit.
An independent second opinion on your architecture, codebase shape or platform direction, with specific risks and recommendations.
Ongoing engineering leadership: technical direction, sound delivery decisions and a steady hand through ambiguity.
Domain Expertise
Built and ran systems that processed payments, transfers and settlement flows, and moved money across borders.
Built and optimised operational tooling that kept support timely and useful.
Designed systems and controls for operational resilience that caught risk early while minimising customer friction.
Worked with auditors through regulatory compliance checks.
Anomaly detection at Wise: £9M+ a year in prevented exposure
Built Making Tax Digital products end to end against HMRC’s APIs, and supported legacy Self Assessment.
Used Open Banking integrations for automated record keeping and MTD reporting.
Experience
Still honing my craft and sharpening my skills, but I wouldn’t be where I am, or half as capable, without these experiences. Grateful for the wins and the mistakes alike.
With Equal Experts, delivered for HMRC, HSBC, O2 Telefónica and Not On The High Street; directly, for Wise and Record OS.
Independent Contractor, via Vicron Tech Ltd
Built MTD capability from scratch, with AI-assistance
This was the first project I delivered through my own company, Vicron Tech. The goal was to build Record OS’s Making Tax Digital (MTD) capability end to end. I delivered the piece mostly solo, from working out the scope through to the API integration and the corresponding frontend.
I also evaluated open-banking partners and built manual bank-statement imports, as a first step towards automated bookkeeping and transaction categorisation.
Entrepreneur in Residence
9 weeks crash course in founding a VC scale company
Antler’s Entrepreneur in Residence was an eight-week programme to get from an idea to a VC-backed company. I was part of their spring 2025 cohort (UK13).
During the programme I met several talented folks from a variety of backgrounds. I formed teams with a few of them, experimented with various ideas from different domains, including insurance, supply chain, finance; pivoted a few times and eventually landed on the venture to simplify the lives of first-time parents.
Other than the domains, this was also a fantastic introduction for me as a technical person to the AI world, and how the software industry was on the precipice of transformation. I ended up building a multi-agent AI prototype with voice and WhatsApp integrations, to not just experiment with AI as tech, but also to explore the new standard of building extensible systems that was already in the making.
Getting a working prototype up was the easy and fun part though. The hard part was everything else - validating the problem, identifying the ICP, researching the market, and most importantly staying in sync as a team.
At the end of the programme, Antler decides which teams go through to pitch to their Investment Committee. We were picked, which meant a ninth week to prepare for the pitch itself.
I voluntarily stepped away partway through that ninth week, before the pitch. The biggest factor was the size of the investment on offer, along with terms I wasn’t comfortable with.
Engineering Lead
Delight grows the business; the boring engineering keeps it alive.
Wise is where I really learned how money moves around the world, and I saw it from both ends. Early on I was out at the edges and mostly on my own, integrating banking partners, launching new currency routes, onboarding customers. In the last few years I moved onto the core platform and stepped up to lead the team there, six senior engineers split between London and Singapore, owning the partner-integration and linking domains that sit under every penny that moved across the company. That’s where the harder problems live: scale, data integrity, and keeping everything reliable and available under real load.
My biggest takeaway from those years was really about product thinking. It’s where I understood that delighting customers - exceeding their expectations, not just meeting their basic needs - is what earns their loyalty and grows the business. You don’t get there by selling people things they don’t need. You get there by solving a real problem so well that they trust you with the next one.
The second lesson ran the other way: the unglamorous, boring engineering is exactly what keeps a business like that alive. Functional matters far more than beautiful, because at this scale the smallest oversight costs real money. A background thread that quietly stops pulling the latest FX rates. A connection pool sized wrong, or one left open a beat too long. A transaction with the wrong scope. Any one of those can bleed millions before anyone notices, and take reputation and trust down with it. So the boring disciplines aren’t optional: throttling, timeouts, retries, failing gracefully without ever corrupting data. Some of my team’s best work came out of that mindset, like the anomaly detection we built to catch that whole class of mistake before it turned into a loss, where I drove the scope and direction.
Leading taught me as much as the engineering did. A lot of it was managing my own time and energy: grooming the people around me, still growing myself, and clearing ambiguity out of the way so the team could stay focused. I had to learn to delegate, which didn’t come naturally at first. I got better at separating the signal from the noise, learned to lean on real cross-team collaboration to get things done with less friction, and came to trust data-driven decisions over gut feel.
Senior Software Engineer
Set the ego aside, and software becomes a craft.
I spent six years here, and honestly the first job was unlearning a lot of what I’d picked up at Infosys. That’s the lesson I took from it: it’s never too late to learn a better way of doing things, and you grow a lot faster once you get your ego out of the way.
It’s also where I learned to build production software from scratch, with confidence: clean code, testing, CI/CD. Software stopped being grunt work and became a craft I cared about. I owe most of that to my mentor here, Mr. Dhaval Dalal, who walked me through the three levels of craftsmanship - apprentice, journeyman, master - and, more than anyone, got me to fall for the work itself.
A few things from those years still run my thinking. You can get real value in front of a customer in days, not months, if you work lean. Agile is a mindset, not just a pile of process. Progress beats perfection. And there’s a genuine difference between building the right software and building the software right. I didn’t develop deep domain expertise in any one business, but I came out able to walk into almost any stack and get moving.
Senior Systems Engineer
Where knowledge turned into practice.
Infosys was my first job out of college. I trained in .NET for five months, spent another five on a mainframes project, and then moved onto a Java, JSP and servlets project that I stayed on until I left in late 2012. Most of it was backend work for a big US healthcare insurer.
More than anything, it’s where my knowledge of software engineering turned into practice. I applied the fundamentals I’d learned to real projects for the first time, got to grips with the basic building blocks of an enterprise system, and learned to adapt quickly as I was moved between very different stacks. It’s also where I made a few lifelong friends, from the group of us who started out together.
Technical Expertise
Projects
Research threads and product ideas I work on alongside client engagements, each at a different stage. If any of them overlaps with your world, I’d love to hear from you.
Show founders and solopreneurs which problems the market has already validated, and where a wedge might be.
A concept for reading VC funding as evidence of validated problems, so a founder’s time goes into solution and distribution instead of validation.
Current question
Can funding flows, read carefully, tell a founder which problem spaces are validated and where the gaps are?
Give parents back the time and calm for the parts of raising a child they’ll want to remember.
A concept for easing the stress of parenting admin: the appointments, forms, deadlines and entitlements that start piling up before a child even arrives.
Current question
Raising a child is a 20-year project; how much of its admin could be anticipated and quietly handled?
If you're a parent buried in the admin, or you've tried to build for this, say hello.
Enable engineering teams to build robust, secure AI-native systems.
Early work toward a simple set of accessible building blocks for robust, secure AI-native systems: evaluation, observability, reliability and day-to-day developer workflow.
Current question
What would make AI systems easier to reason about, test and operate once they are part of real software delivery?
Running AI in production and finding it fragile? I'd like to compare notes.
Take the overwhelm out of HMRC and open-banking integrations so small teams can ship them faster.
An attempt to give teams a structured, incremental way into Making Tax Digital and open-banking integrations, with reusable building blocks informed by what I learned delivering Record OS’s MTD capability.
Current question
Which parts of a regulated integration can genuinely be reused across firms, and which are always bespoke?
Contact
For AI workflows, product ideas, technical questions, or a problem you’re weighing up - big or small, start with a little context.
As of July 2026: available for new engagements.
A 30-minute conversation to walk through your situation and work out the most useful next step.
Book a timePrefer to write first? Share the context, the constraint and what would make the conversation useful. I reply within one business day.
Thanks — received. I’ll be in touch.
You can also reach me by email at [email protected] or connect on LinkedIn.