Callum
Lindsey

AI engineer.

I build things
end to end.

I'm an AI engineer at a large UK tech company, building multi-agent systems and automated workflows from first prototype through to production.

UKLinkedIn Skip to experience

Highlights.

What I bring
to a team.

Agent workflow: idea → production
Prototypes in a couple of hours
Always on the latest AI tools
My team’s go-to for AI tooling
Graduate to AI engineer in under 4 years

AI means I can fail fast and succeed fast: tools and workflows that add value, built faster than ever and to the same quality.

I’m an AI engineer in the UK. I build multi-agent pipelines, internal AI tools and automated workflows, and take them from a prototype in a couple of hours through to production. I got into the role by automating parts of my job for the team, going from graduate scheme to AI engineer in under four years, and I’ve always loved trying the latest tools and tech. Before that: test automation, Dynamics 365 development, and five years in retail management alongside a First in Business IT.

What I’m good at: Finding the real problem and fixing it with a full solution. Picking the right tool for the cost. Explaining AI plainly, so people actually use what I build. And delivering when I say I will.

Outside work: cat dad to Teddy, and always tinkering with the home lab and other projects.

Experience#

At a large UK tech company

2026 to now
AI Engineer
Build multi-agent pipelines, internal AI tools and automated workflows, such as a support agent that triages issues and works with other AI agents to fix them. Prototypes in a couple of hours, then on to production.
- Cut prototype turnaround from weeks to a couple of hours.
- Rebuilt the same agent pipeline three times as cost, constraints and models changed, most recently as a production version others can use.
- Shipped a self-serve requirements tool used across the department and other teams: it questions vague requests until the real need is clear, then feeds the pipeline.
- Built a retrospective tool that pulls sprint data in over MCP and uses AI to surface points worth raising.
- Started the design-system pipeline from Figma to Storybook, packaged as a component library now used across the department.
- Evaluated replacement models against real pipeline runs, set which model each agent uses, and shared the cost approach with other teams.
- Helping architect the platform the pipeline runs on: a .NET 8 API triggering Temporal workflows with Python workers.
- Wrote two Model Context Protocol servers from scratch to automate my own workflow, self-hosted internally.
- The person colleagues in my team come to for AI tooling.
Claude Code · Codex · Pi · OpenCode · Roo Code · Pydantic · Temporal
2023 to 2026
Test Automation Engineer
Moved a BDD test suite to Python and Playwright, and started the AI work that led to my current role: building orchestrators in Roo Code to automate day-to-day tasks.
- Tested web, mobile web and APIs with Bruno and Postman.
- Worked with non-technical colleagues to turn what they needed to be sure of into scenarios that could actually be run.
Python · Playwright · BDD with Gherkin · Roo Code
2023
Junior Dynamics 365 Developer
Configured and extended Dynamics 365 CRM and Power Apps for internal business users, and evaluated new features for the value they could add.
2022 to 2023
Graduate
Rotated through each stage of how the department delivers its systems, from requirements to live support:
- Business analysis: gathered requirements, mapped processes and wrote acceptance criteria.
- Development: built features and fixed bugs.
- Testing: manual and regression testing, and my first test automation.
- Application support: worked Dynamics 365 support tickets with internal teams, and managed fixes with a third-party supplier.
Before that
2017 to 2022
Specialist and Duty Manager, Halfords
Ran the store day to day as Duty Manager, and a department as Specialist, alongside full-time study: exceeding sales and service targets and training colleagues.

Other builds#

Things I’ve built outside work.

Multi-channel AI assistant
Web chat, WhatsApp and voice from one knowledge base, with a loop that logs unanswered questions so the gaps get closed.
n8n · Supabase · pgvector
Plan-reading estimator
A quantity-extraction tool using AI agents to read dimensions straight from vector PDFs, benchmarked and tuned against a hand-labelled gold set.
AI agents · benchmarking · tuning
Sites and home tools
Websites, including this one, and small tools for home.
Astro · Three.js · Cloudflare

Education and skills#

Education

BSc (Hons) Business Information Technology, First Class

Staffordshire University

2017 to 2020

Enterprise cloud (AWS, Azure, GCP), data and SQL, and Python. Final project: a costed cloud migration plan for UK education IT, using Raspberry Pis to bring the costs down.

AI and agents: multi-agent orchestration and agent design, Model Context Protocol servers and clients, Claude Code and OpenCode (skills, hooks, subagents), Codex, Pi and Roo Code, Pydantic, prompt engineering, model evaluation and selection, AI-assisted development
Languages: Python, C# / .NET 8, SQL, Bash
Build and infrastructure: Docker, Linux, Git, GitLab CI, Temporal, n8n, Supabase / PostgreSQL, Google Cloud and GKE, AWS
Testing: Playwright, BDD with Gherkin and SpecFlow, API testing (Bruno, Postman), LambdaTest, manual, exploratory and load testing
Platforms: Dynamics 365 (CRM), Power Apps
Analysis: requirements gathering, process mapping, acceptance criteria
Home lab: Home Assistant, Docker, UniFi and VLANs, Pi-hole, Arduino

Home lab#

Where new tooling gets broken first.

TestbedRunning
A Dell box running Linux and Home Assistant. The household uses it daily for lights, notifications and routines.
NetworkRunning
UniFi kit split into VLANs, with Pi-hole blocking ads for the whole house.
Bin lightRetired
Changed colour by bin type. Retired once the council calendar could do the job.
LED boardRetired
Scrolling text and pixel art. Retired when its power supply went bang.

Read the write-up