Skip to content

Forward deployed AI engineer

AI that worksinside your business.

I’m Roy. I build AI assistants, automate workflows and connect AI to the systems your teams already use.

A hands-on engineer and engineering manager, working with your people from the first useful idea to a system they can rely on.

Let’s discuss your projectMeet the engineer behind it
  • Defined projects
  • Embedded delivery
  • Fractional leadership
22+ years
building software
Production AI
LLM & retrieval implementations
Engineering leadership
teams, platforms & delivery

What I can implement.

Where could AI make work better?

Bring me a slow process, scattered knowledge or an AI idea you need to deliver. I’ll help define the opportunity and build the right solution with your team.

  1. Give your team answers they can use.

    Internal AI assistants that find and explain information across your documents, support history and business systems, with source references and access controls.

    An answer, with sources
    • Internal knowledge search
    • Support assistants
    • Document Q&A

  2. Move repetitive work forward.

    Workflows that read incoming information, extract what matters, prepare the next step and update your tools. Add human approval wherever the decision needs it.

    Then your tools are updated
    • Document processing
    • Request triage
    • Approval workflows

  3. Build AI into your product.

    AI features inside the software your customers already use. From a working prototype to an integrated experience, with evaluation, monitoring and a clear rollout plan.

    Designed around your users
    • Product assistants
    • Content workflows
    • AI-powered search

  4. Connect the pieces. Make it dependable.

    Connect AI to your APIs, CRM, CMS and operational tools. Put permissions, reliability, logging and ownership in place so your team can run and improve the system.

    Permissions, logging and ownership built in
    • API integrations
    • Production readiness
    • Technical leadership

Forward deployed, in plain English

I get close to the problem. Then I build.

I work directly with the people doing the work, understand the systems around them and write the code to improve the process. You get someone who can talk to the business, make technical decisions and lead delivery through to adoption, and who is comfortable where requirements are ambiguous, the data is sensitive and the stack has history.

It’s the operating model the leading AI labs use to close the gap between what a model can do and what a business needs. Whatever gets learnt in the field is codified into evals, playbooks and tooling your team keeps.

Tell me what you’re working on

  • Find a useful starting point

    Understand the workflow, data and constraints. Agree what success looks like and whether AI is the right tool.

    Outputa clear problem and a scoped delivery plan

  • Build with real users and data

    Connect the relevant systems, create a focused first version and prove it against an evaluation set drawn from representative work with your team.

    Outputa working implementation and the evals that prove it

  • Ship it. Help your team own it.

    Roll out carefully, monitor behaviour and support adoption. Document the system and transfer the knowledge to your team.

    Outputa production system and a practical handover

About me

The person you’d be working with.

I’m Roy Barber, an engineer based in Doncaster in the UK. I’ve spent 22 years building for the web: starting out freelance, founding and selling a digital agency, co-founding an education platform and leading engineering for enterprise SaaS.

These days the work is applied AI: taking LLM and retrieval systems into production and helping teams actually adopt them. The thread through all of it is the same: useful, accessible software that people rely on, built close to the people who use it.

I’ve worked remotely for 12 years and run teams the same way. Every week my current team sets time aside for what we call Ferrari Friday: we stop, reflect and work out how to make the car go faster. I bring that habit of deliberate improvement to every engagement.

Portrait of Roy Barber

The experience behind the work

Hands-on engineering. Business perspective.

I’ve built products, founded and sold an agency, integrated financial systems and led engineering teams. I bring that delivery experience to practical AI implementation.

Full background on LinkedIn
Companies I have worked with: Google, Castrol, UNICEF, Karndean Designflooring, Nestlé, VolvoDacia, Strata, Solicitors Regulation Authority, Supercar Driver, The Wine Society, Keepmoat

  • Happydance

    Applied AI & engineering leadership

    Head of Engineering, 2024 to present

    Lead the function spanning engineering, product, infrastructure and QA. Set direction for enterprise recruitment software and integrate LLM and retrieval capabilities into production products and workflows.

    From AI implementation to team delivery.

  • Umbraco

    Headless CMS & developer community

    Headless Community Team, 2022 to present

    Part of the community team shaping Umbraco’s headless CMS: advising on APIs, content structures and developer experience for modern JavaScript frameworks.

    Helping shape the platform, not just building on it.

  • Dotcentric

    Enterprise platforms & delivery

    Head of Frontend, 2021 to 2024

    Led the frontend engineering team delivering enterprise headless CMS and e-commerce platforms, accessible to WCAG 2.2 and recognised with industry awards.

    Enterprise delivery with accessibility as standard.

  • Roy&Co

    Commercial delivery & ownership

    Founder & Director, 2014 to 2021

    Founded and later sold a digital agency. Built engineering teams and led client relationships and full-stack delivery for organisations including Google, GoCompare, Castrol and Co-operative Insurance.

    An engineer who understands the client’s side of the table.

Built, founded & sold

Skin in the game.

I’ve built a product through to acquisition, and founded, run and sold businesses of my own. It shapes how I build for yours: with an owner’s view of cost, risk and what actually moves the business.

  • AppToPay

    Consumer finance platform

    I built the product: a regulated consumer finance app with credit-scoring and identity-verification integrations and the customer onboarding journey.

    Acquired by Hitachi for £2.4m

  • eduly®

    Education platform · Nov 2017 to Apr 2021

    Empowering schools, teachers and parents. eduly® handled a primary school’s communication, consent and engagement requirements in one central platform.

    Sold for £1.3m

  • 180 Degrees Fitness

    Boutique gym · May 2016 to Jun 2019

    Retford’s most technologically advanced boutique unisex gym, run as an owner-operator alongside my engineering work.

    Sold for £500k

Freelance & contract engagements

The right support for the work ahead.

Bring me in for a defined project, embed me with your delivery team or add senior engineering leadership on a fractional basis.

A few useful details

Before we talk.

We have an idea, but no specification. Is that enough?

Yes. Start with the process you want to improve, who uses it and what is difficult today. We can work through the opportunity, dependencies and a sensible first scope together.

Can you work with our existing engineering team?

Yes. I can own a piece of implementation, work alongside your engineers or provide technical direction and coaching. We agree responsibilities, access and handover at the start.

What happens to our data and permissions?

We review the data involved, who should have access and which providers and systems are suitable before implementation. The solution is designed around those decisions, including approval steps and access controls where needed. This isn’t new ground for me: I’ve built consumer credit onboarding with identity verification, delivered for insurance and financial services brands, and handled children’s data in education and candidate data in enterprise recruitment.

How do we know an AI implementation is useful?

We agree measures around the actual workflow, such as time spent, answer quality or review effort. Then we test representative cases with your team and use those results to decide what is ready for rollout.

Can you work alongside an AI vendor or platform team?

Yes. As well as working directly for a business, I take engagements with AI product companies that need forward deployed capacity with their customers: discovery, deployment and evaluation work in the field, with what gets learnt fed back to the product team.

Let’s put AI to work

What would you like to make happen?

Tell me what your business needs, the systems involved and where you’re getting stuck. We’ll work out whether I’m the right person to help.

hi@roybarber.com