About
I'm Dan. I build software, and lately the systems that let agents build it with me.
I'm a Lead Software Engineer based in Bolton. I lead a team building cloud-native, serverless applications in TypeScript on AWS, and I care about the unglamorous parts: system design, automated tests and releases that don't need a meeting.
These days most of my attention goes on one question: how do you get real, shippable work out of AI agents without becoming the bottleneck yourself? This site is my notebook for the answers that held up, written as methods you can lift into your own team, not as tool reviews.
- 2024–now
Lead Software Engineer · ATG Entertainment
Lead the team building serverless applications in TypeScript on AWS; set technical direction and mentor senior and mid-level engineers.
- 2023–24
Senior Software Engineer · ATG Entertainment
Owned delivery of microservices end to end; pushed CI/CD and infrastructure as code.
- 2022–23
Software Engineer · ATG Entertainment
TypeScript and Node.js services.
- 2020–22
Web Developer, then Team Lead · Product Care Group
Serverless, event-driven microservices: an asynchronous order system on Step Functions, DynamoDB, Lambda and SQS, with React front ends on Amplify.
- 2019–20
Junior Software Developer · 3P Logistics
Internal systems, customer dashboards and mobile apps.
- 2018–20
Founder · Bee Design Studio
Websites and hosting that got local small businesses online.
- 2007–20
DJ and promoter · Bolton
Ran club nights for years, posters and all. Good training for reading a room.
Education
MSc Computing (Software Engineering)
The Open University · Distinction
160 benchmarks later…
Dissertation
Performance and cost-optimisation of serverless architectures: scalability, orchestration strategies and execution efficiency.
- a.512 MB is the inflection point for CPU-bound Lambda work. For I/O-bound work, more memory only multiplies the bill.
- b.Orchestration has a routing tax of around 1.5 seconds through EventBridge and Step Functions, compared with direct invocation or queues.
- c.Provisioned concurrency can cost more when traffic overflows it and spills into on-demand billing.
The result was a decision framework for choosing an orchestration pattern by latency, cost and carbon footprint.
Try the decision tool →What I write about
- 01Engineering with AI agentsStructuring agent workflows, keeping people at the decisions that matter, and running agents safely in production.9 posts →
- 02Serverless on AWSLambda, DynamoDB, EventBridge and infrastructure as code: event-driven back ends that stay cheap and boring.10 posts →
- 03TypeScript, end to endOne language from the API to the interface, with types doing the work that tests would otherwise have to.4 posts →
How I work
- Agents do the legwork; people make the calls that are expensive to undo.
- Cut what doesn't ship. Every agent, feature and dependency earns its place.
- Boring beats clever: append-only files before queues, one language before three.