Answers
Direct answers to the questions founders, engineers and technical leaders ask about AI engineering, technical co-founders and CTOs, and building AI products in the UK. Each page puts the answer first, then the facts with their sources, written by Dr. Dulan Dias, co-founder and CTO of Tellme AI and author of From Model to Product.
The book and AI engineering careers
Who is Dr. Dulan Dias?
Dr. Dulan Dias is a Belfast-based technology founder and AI engineer. Dulan is co-founder and Chief Technology Officer of Tellme AI; founded EchonLabs in Sri Lanka in 2017 and grew it to a team of 60+ before its acquisition in 2026; holds a PhD in Artificial Intelligence and Data Science from Queen's University Belfast; and wrote From Model to Product, a 2026 book on taking AI from prototype to production.
What is From Model to Product about?
From Model to Product: Foundations for Practical AI Engineering, by Dr. Dulan Dias, is a 403-page book about turning AI models into dependable products. Its argument is that the hard part of AI is not the model but everything around it: problem framing, data, APIs, storage, deployment, monitoring, reliability, cost, teams and governance. The book covers back-end and production engineering; it does not cover front-end work or agentic AI.
What are the best books on AI engineering?
The best books on AI engineering in 2026 are Chip Huyen's AI Engineering for building on foundation models, Designing Machine Learning Systems by the same author for ML system design, Hands-On Large Language Models and Build a Large Language Model (From Scratch) for how LLMs work, and Reliable Machine Learning and From Model to Product (by this site's author) for running AI in production. Start with the one that matches your gap.
What is AI engineering?
AI engineering is the discipline of building, shipping and running software products whose behaviour depends on AI models, today most often foundation models used through APIs. The core of the job is the system around the model: framing the problem, choosing the approach, assembling context, evaluating outputs, and keeping quality, cost, latency and reliability under control. AI engineers usually adapt existing models rather than train new ones.
How do you become an AI engineer from software engineering?
A software engineer becomes an AI engineer by adding skills in this order: calling models reliably, evaluating their outputs, grounding them with retrieval, enough classical machine learning to choose the right approach, and running AI features in production, with fine-tuning and agents last. Prove each skill with a small shipped project, not a certificate. Allow roughly six to nine months of part-time study; that is a worked estimate, not survey data.
Technical co-founders and CTOs
What does a technical co-founder do?
A technical co-founder owns a startup's technology as a founder: they decide what to build and how, build the first version largely themselves, hire and lead the first engineers, and answer for whether the product works, stays secure and can grow. Unlike a first engineer, they share the founders' risk and equity and are accountable for outcomes, not tasks. The job changes at every stage, from builder to engineering leader.
Do I need a technical co-founder, now that AI can write code?
You need a technical co-founder when technology is what your company sells or the reason it can win: deep tech, AI-native products, anything investors will judge on its engineering, and products that must keep changing for years. AI coding tools, an agency or freelancers are enough to validate demand, run a service business on software, or build tools that support a non-technical business. AI writes code; it does not own the product.
How do you find a technical co-founder in the UK?
The best place to find a technical co-founder in the UK is among engineers who have already seen you work. Beyond your network, the structured routes are talent investors such as Entrepreneur First and Antler in London, YC Co-Founder Matching, PhD venture programmes such as Conception X, university communities and technical events. Lead with evidence of the problem, not the idea, and test the partnership on a short real project before agreeing equity.
How much equity should a technical co-founder get?
A technical co-founder who starts with you and commits full time should usually get an equal or near-equal share of the founder equity, because the value lies in years of execution, not the idea. Less is fair when they join after the product or funding exists, take a full salary or put in no cash where others do. Whatever the number, put it on vesting with leaver terms, and take UK legal and tax advice first.
What should you ask a potential technical co-founder?
Ask a potential technical co-founder questions that make them show how they decide, not what they know: when their last system failed and why, what they would refuse to build, how much time they can commit and when, what they need to earn, and how they handle being overruled. Good answers are specific, admit trade-offs and mistakes, and connect technology to customers. Rehearsed certainty and vague commitment are the warning signs.
Technical co-founder vs CTO vs fractional CTO: what's the difference?
A technical co-founder, a CTO and a fractional CTO differ in what they own. A technical co-founder owns part of the company and builds it from the start, usually for founder shares and a low salary. CTO is a job title: the executive accountable for technology, held by a co-founder or a hire. A fractional CTO is a part-time senior leader paid by the day or month, who owns agreed outcomes rather than the company.
What does a CTO do at an AI startup?
A CTO at an AI startup does the usual CTO job, owning architecture, the engineering team and technical risk, plus decisions most software CTOs never face: whether to build, buy or host models, what data the company may use, how quality is evaluated, what each user costs in model calls, and how the product fails safely. The role is judged by whether the AI works dependably for customers at a cost the business can carry.
How do you build an AI team at a startup?
Build an AI team at a startup by hiring for the product, not the model. Start with one or two engineers who can ship an AI feature end to end, involve a domain expert from the first week, and add specialists in data, machine learning or infrastructure only when a measured bottleneck appears. Most early AI products need software and data engineering more than research, so machine learning specialists usually join after the first data and evaluation results.
How do you scale an engineering team from 5 to 50+?
Scale an engineering team from 5 to 50+ by changing structure, process and leadership as communication load grows, ideally just before it forces you to. Split into small teams that each own a product area end to end, add engineering managers when leads can no longer both build and support people, write decisions down, and create platform teams once several teams repeat the same infrastructure work. Measure delivery as you grow, so you know whether each change helped.
What do investors check in technical due diligence for an AI startup?
Investors use technical due diligence on an AI startup to check three things: that the product works as claimed, that the company owns what it sells, and that it can grow without technical surprises. Expect questions on architecture, code quality, security, rights to training and customer data, dependence on model providers, cost per user, evaluation evidence, key-person risk and IP assignment. The depth grows with the round, from a conversation at pre-seed to a structured review later.
Building AI products
What is agentic AI?
Agentic AI is software in which a language model decides its own next steps: it calls tools, reads the results and keeps going until a goal is met, instead of following a path a developer fixed in advance. The useful dividing line is control. In a workflow, code chooses the sequence; in an agent, the model does. Agents suit open-ended tasks with checkable outcomes, and they cost more, run slower and fail in compounding ways.
RAG vs fine-tuning: which should you use?
Use retrieval-augmented generation (RAG) when the model needs knowledge it lacks, especially facts that change, are private or must be cited. Use fine-tuning when the model has the knowledge but behaves wrongly: the wrong format, tone or consistency at a narrow task, or a large model doing a job a smaller tuned one could. Try prompting first, then RAG, then fine-tuning, and combine them when you need both.
How do you evaluate an LLM application?
Evaluate an LLM application with a versioned set of real test inputs, graded by code checks wherever possible and by rubric-based LLM judges, calibrated against human ratings, where not. Run it on every change to prompts, models, retrieval or tools, then keep watching production, because offline scores only predict live quality when the test set looks like real traffic. Measure the whole system, not the model alone.
How do you reduce LLM hallucinations in production?
Reduce LLM hallucinations by giving the model the facts instead of relying on its memory, retrieving only relevant context, constraining outputs with schemas, requiring citations that code can check, allowing the model to say it does not know, and sending high-stakes answers to a person. Hallucinations can be reduced but not eliminated, so design the product to detect and contain the ones that get through.
Should you use an open-weight model or a proprietary LLM API?
Start with a proprietary LLM API unless you have a specific reason not to: it gives strong quality quickly with no infrastructure to run. Choose an open-weight model when data must stay in your environment, volume is high and steady enough to keep hardware busy, you need to fine-tune or keep a model version indefinitely, or a smaller model meets your quality bar. Read the licence of the exact model first.
What is an AI-native company?
An AI-native company is one whose core product could not exist without AI: remove the model and there is nothing left to sell. An AI-enabled company adds AI features to a product that already works without them. The difference shows up in four places: what the product promises, how data flows back into it, who the team hires first, and unit economics, because every use carries a model cost that traditional software never had.
What tech stack should an AI startup use in 2026?
For most AI startups in 2026 the right stack is a boring one: a TypeScript web front end, a Python API such as FastAPI, hosted models called only through your own backend, PostgreSQL with pgvector for data and retrieval, a queue for slow work, tracing that records tokens and cost, and one cloud region close to your users. Deviate only when a measured constraint such as latency, data residency or cost at volume forces it.
How much does it cost to build an AI MVP?
The cost of an AI MVP is mostly people time: the engineer-months it takes, multiplied by what each month costs you. Model fees and hosting are usually a small share at MVP volumes, because a pilot makes thousands of model calls, not millions. In the worked examples below, an API-based MVP comes to about £47,000 and a custom-model MVP to about £153,000, on stated assumptions you should replace with your own.
What AI regulation applies to UK startups?
As of November 2026 the UK has no general AI law. A UK startup that builds or uses AI is regulated through existing law: UK GDPR and the Data Protection Act 2018, as amended by the Data (Use and Access) Act 2025, plus equality, consumer and sector rules enforced by regulators such as the FCA, MHRA and Ofcom. Selling into the EU can bring the EU AI Act into play. This is general information, not legal advice.
How should a UK small business start using AI?
A UK small business should start with one costly, repetitive problem, not with a tool. Check that the data the problem needs is usable and lawful to use, try an off-the-shelf product before building anything, run a time-boxed pilot against a measured baseline, then roll out with simple rules on data and human review. As of November 2026, the UK help that exists is mainly free training, Knowledge Transfer Partnerships and local business support.