About
A small practice that ships production AI
Alkemia Technologies is a London AI engineering practice. We build systems that do work inside your business, not demos that impress in a meeting.
Founded in 2019, we started as a software practice building e-commerce and SaaS products for London businesses. That work is still here — but since 2023 the centre of gravity has moved to AI engineering, because that is where our clients’ hardest problems moved.
What we do is narrower than most agencies advertise. We build tool-calling agents that take real actions behind approval gates, retrieval systems that cite their sources, and the evaluation infrastructure that tells you whether either is working. We are equally willing to tell you that a problem does not need AI at all.
6+
Years shipping production software
13
Tools in one shipping agent
7
Models with fallback routing
5M+
Records in one product data layer
How we build
Positions we have arrived at
AI systems fail in ways ordinary software does not. These come from shipping them.
Write actions need gates
An agent that can only read is a search box. An agent that can write needs enforcement in your backend — not an instruction in a prompt asking it politely to confirm first.
Evaluation before scale
If you cannot measure whether a change made the system better, you are not iterating, you are guessing. We build the measurement first, even when it delays the demo.
Model-agnostic by default
Providers change prices, deprecate models, and have outages. Model choice belongs in configuration with fallback ordering, so switching is an afternoon rather than a rewrite.
The boring parts decide it
Auth, permissions, billing, idempotency and error handling are what separate a convincing prototype from something you can put in front of customers. Most AI projects die there.
Cost is a design constraint
Route cheap work to cheap models and reserve frontier models for where accuracy pays for itself. Meter it, so the bill never arrives as a surprise.
Say when AI is the wrong tool
Plenty of problems are better solved by a query, a rule, or fixing a process. We would rather tell you that during an audit than discover it together six months in.
Who you work with
Ozan Yildirim
Founder & AI Engineer
Alkemia is deliberately small. You work directly with the engineer building your system — there is no account manager between the conversation and the code.
Ozan holds an MSc in Computer Science from ESIEE Paris, specialising in machine learning and deep learning, on top of a BSc in Mathematics. Before founding Alkemia he worked in data science at IQVIA’s Analytics Centre of Excellence in Paris and as a developer at UniCredit Bank.
Since 2019 he has shipped production software for logistics, print and hospitality businesses across London, and builds and sells his own AI products — which is where a good deal of the sharper engineering gets worked out before it reaches client work.
Next step
Tell us what you’re trying to build
Most engagements start with a fixed-price audit, so the first thing you buy is a decision rather than a commitment.