Engineering lead · Technical Architect · Fractional CTO · London

AI writes code.
Someone still has
to decide what.

I architect agentic systems, direct AI-driven development, and own what ships. The judgment, the scoping, the accountability — that's the part that doesn't get automated.

17 years across DAZN, EF Education First, MongoDB, Satalia and a dozen others. £600M in revenue enabled. Available for new contracts from January 2027.

Start the conversation

+12

17 years. Start-ups to global enterprises.

Companies I've shipped for

The tools changed.
The judgment problem didn't.

Your team can generate code faster than ever. What they can't generate is the experience to know whether what they're building will survive production, scale, or the requirement they haven't thought of yet.

I work with Claude, Cursor, and agentic tooling as a force multiplier — designing the system, setting the constraints, reviewing what the agents produce, and catching what junior engineers miss. The output is faster. The accountability is still mine.

This isn't about being sceptical of AI. It's about knowing which decisions require a human who has watched systems fail at scale — and making sure those decisions get made correctly.

You've adopted agentic tooling but no one's steering it

AI agents follow instructions precisely. I make sure the instructions are right before anyone types them.

Your AI-generated codebase is growing faster than your confidence in it

I audit what's been built, identify the structural risks, and put the right guardrails in place.

You're migrating legacy systems to AI-native architecture

17 years of knowing what's worth keeping — and what will quietly break everything if you don't replace it first.

You have a launch in weeks and the current build won't survive it

Some things haven't changed. I've rescued three products days before go-live. I know what to cut.

Agentic systems architecture

I design the systems that AI agents build inside. Constraints, interfaces, review loops, failure modes — the scaffolding that makes agentic development safe at scale.

AI integration that ships

Enterprise AI wired into production infrastructure — OpenAI, Claude, MLOps pipelines — secure, auditable, and built to survive the next order of magnitude.

Fractional CTO

Technical leadership across the full picture — architecture, agentic tooling strategy, team structure, board-level communication. Personal. No handoffs.

Case study

£250k in month one — from a prototype that couldn't launch.

TOCA Social had a booking platform, a launch date two months out, and a prototype that wouldn't survive real traffic. They needed someone to take ownership — not manage it.

I reviewed the existing build, presented a plain-English report on what needed to change, and rearchitected the entire platform on AWS serverless — Terraform, Node.js, SNS/SQS, DynamoDB, Docker. We launched on time. 3,200 bookings. £250k in the first month.

8 wks

diagnosis to launch

£250k

revenue, month one

3,200

bookings at launch

What was broken

— Prototype built for demo conditions, not real load
— No scalable booking or scheduling logic
— No DevOps, no deployment pipeline, no observability
— 8 weeks to a public launch with paying customers

What shipped

— Full serverless AWS rebuild from scratch
— Real-time booking and availability engine
— Terraform-automated infrastructure and CI/CD
— Launched on schedule, zero downtime

Selected results

DAZN · Live streaming

£600M

annual revenue enabled

Global streaming API serving 30k requests per second at 99.99% uptime. Built the initial microservices, supported the release, and drove the scalability work that made this revenue ceiling possible.

Belron · Field operations

2×

technician daily productivity

Real-time 'Uber-style' scheduling system for field technicians, customers, and operators. Reduced human error, improved customer experience, and doubled daily output — rolled out across 4 countries.

MongoDB · AI modernisation

40%

faster enterprise AI adoption

Brought in to drive OpenAI solution innovation and accelerate enterprise cloud migration. Strategic technical leadership across modernisation and AI integration workstreams.

Satalia · MLOps platform

1M+

AI workflows in production

Designed AI-ready enterprise platforms across AWS, Azure, and GCP — powering model training, creative analytics, and MLOps operations at scale for global enterprise clients.

Every engagement is led personally by me.

Independent contractor since 2014. Before that, a decade building software at NTT Data, 3Pillar Global, and as a freelancer. I've shipped at every scale — from a two-person startup to a system serving millions of concurrent users across three continents.

The thing I've learned is that most technical failures are actually judgment failures — someone made a decision early that nobody questioned, and it compounded. AI makes this faster in both directions. Agentic tooling amplifies good architectural decisions and catastrophic ones equally. My job is to make sure the right decisions get made before the agents start building.

I use Claude, Cursor, and agentic development tooling as a core part of how I work. Not as a shortcut — as a force multiplier that lets a small, senior team move at a pace that wasn't possible three years ago. The accountability for what ships is still mine.

"AI writes code faster than ever. Someone still has to decide what to build, catch what it misses, and own what ships."

Alex Ionica · Founder, TheDev.ai LinkedIn →

ALEX

Available from January 2027

Planning something for the new year?

The best engagements start with a conversation weeks before the work does. Tell me what you're building, what's at stake, and when you need to move — I'll tell you whether I can help.

No pitch decks. No account managers. A direct conversation with Alex, now, so you're not scrambling in January.

Email Alex directly

alex@thedev.ai

AI writes code.Someone still hasto decide what.

The tools changed.The judgment problem didn't.