ENDEWhy this person.
Most AI projects fail for one of two reasons. Either the person who understood the business could not build, so the thing that got built solved the wrong problem. Or the person who could build did not understand the business, so it solved a problem nobody had. Almost nobody sits in both chairs, which is why the same project keeps failing in the same two ways.
Dinobridge is owned and led by Moha Aghanoori. This page is the case for why he sits in both.
The engineering record.
-
MSc Artificial Intelligence
A full master’s degree in the subject, not a weekend certificate.
-
AI Solutions Engineer, Lovehoney Group
Sole engineer on an end-to-end contract lifecycle platform for the group’s legal team: conversational AI intake, an in-app agentic copilot, and a Microsoft Word add-in producing native tracked-change redlines. It replaced a fragmented email-and-spreadsheet process and a commercial vendor costing about €70,000 a year. Every AI action calls the application’s own endpoints as the signed-in user, so it inherits the same permissions and the same audit trail as a human request, behind a confirm gate that pauses every change for approval. Core engineer on a customer-service assistant live across six markets, resolving about 45% of first-line contacts without a human; owner of the retrieval layer and of the prompt-and-model deployment layer the team’s AI tooling ships through.
-
Artificial Intelligence Engineer, Fraunhofer FOKUS
Two years inside one of Germany’s applied research institutes. It is where the research-group work on this site comes from: what a lab’s week actually looks like, why a supervision record matters more than a dashboard, and where the science has to stay human.
-
Machine Learning Engineer, Teck Resources
Production machine learning at industrial scale for a listed Canadian mining company. The first of the large-organisation years, and the one that taught the difference between a model that works and a model somebody else can run.
Why that combination is the qualification.
-
AI engineering, not prompt-writing.
Retrieval pipelines, agents, model and prompt versioning, deployment, and evaluation harnesses that decide what is allowed to reach production. The systems above are live and other people depend on them.
-
Business operations.
Years of it inside organisations where the AI had to work for colleagues who never asked for it, survive a handover, and answer to an auditor. That is a different discipline from making a demo, and it is the one that decides whether a system is still in use a year later.
-
Regulated workflows.
Contracts under confidentiality, with permissions and an audit trail as first-class requirements rather than a later phase. It is why the law-firm work starts with “who did this, and can you show me?” instead of with a model.
-
Leadership.
The people who commissioned those systems were the people accountable for the function. Explaining a trade-off to a decision-maker in their own vocabulary — and being told no — is a practised skill here.
-
Implementation.
Every system described on this site was built by the person who sat in the room where the problem was described. There is nobody to hand it to.
The honest part.
Moha Aghanoori has a full-time job. He is an AI Solutions Engineer at Lovehoney Group in Berlin, and that role is where most of the production experience on this site comes from — it is the reason we can tell you what survives an audit rather than guessing. It also means capacity is real and limited, and you should know its shape before we talk rather than after.
- We take a small number of clients at a time. The line at the top of the home page says whether there is room.
- The shape of an engagement — which days, on site or on video, over how many weeks — is agreed with you in writing after the first call and before there is a contract. We will never tell you a day is available and then move it.
- The Head of AI role is a named window every week, agreed before we start.
- Client work and employed work are kept strictly separate. No client’s material has ever touched another engagement, and none ever will. The systems built as an employee belong to that employer and are described here as employed work, not as client work — because the distinction matters, and because you would find out anyway.
- If what you need is more than that, you hear it before there is a contract. It has happened.
What he will not do.
Build the button that takes the decision away. Automate anything where your name, your licence or your signature is on the result. Sell you a system when a calendar, a form or a shared spreadsheet would fix it — that has cost projects, and it is most of the reason clients introduce us to other owners.
Tell us what your team’s week looks like.
Thirty minutes, free, no preparation. We tell you honestly where AI could change your business, and where it would not.