Enterprise IT Executive and Infrastructure Architect based in South London, with 35 years of experience scaling infrastructure for some of the most technically demanding organisations in the world - Google DeepMind, Shazam, and Apple.

I'm currently between roles after a redundancy and using the time to build things I've always wanted to build, sharpen my Python, and start learning Go.

A multi-repo personal weather station project built around a WeatherFlow Tempest station mounted on my house in Selhurst, South London.

A Python CLI tool that intelligently backs up a developer's projects directory. Scans for project types (Python, Go, Node etc.), checks each has a remote git origin, excludes venvs and build artifacts automatically, creates a manifest with metadata, and supports multiple backup destinations (local, Dropbox, Google Drive, Nextcloud) via a config file. Designed to be useful to any developer, not just me.

Planned

A self-hosted job search tracker built as a Flask web app with a dark theme, multi-user support, and a full contacts/organisations CRM. Tracks every application through its complete lifecycle with a timeline of events, not just a status field.

Designed to answer questions a spreadsheet can't: which sources convert best, how long does the average response take, which agencies are actually performing, and what needs following up today.

Features:

• Full pipeline view with funnel metrics and conversion rates
• Organisations and contacts CRM — track agencies, companies and individual recruiters separately, with contacts shared across applications
• Complete event timeline per application — every email, call and interview logged
• Source intelligence — differentiate direct, agency, LinkedIn, referral and headhunted
• Follow-up reminders and cold application detection
• Deployable as a Docker container for anyone to self-host

Pausedi - I got Capacities to bend to my will and use that for tracking my job applications. Writing a job tracking system when you should be looking for a job is a distraction.

A data analysis tool built on the complete history of UK National Lottery draw results (publicly available as CSV). Finds genuinely interesting patterns in the data — numbers drawn more often on certain days, numbers appearing every March but never in July, longest gaps between appearances, most common pairs and triplets. Not a prediction tool — just honest interrogation of a large dataset. Includes a natural language query interface powered by the Claude API.

Abandoned I was looking for projects with large amounts of data and this seemed an abvious choice but the Lottery company no longer publish all the result.

A Python CLI tool that automatically gathers evidence for common ISO 27001 Annex A technical controls from a Linux system - disk encryption status, password policies, open ports, running services, user accounts, sudo access, failed login attempts and more. Produces a structured report mapped to ISO 27001 controls. Built by a certified ISO 27001 auditor with 35 years of enterprise IT experience.

Planned

Using a Google Coral USB TPU accelerator connected to the homelab server to run real-time ML inference on the rooftop weather camera feed. Classifies sky conditions (clear, partly cloudy, overcast, stormy) directly from the image and cross-validates against the clear sky index calculated from solar radiation data.

Planned

Uses the TfL open API to analyse historical delay patterns and predict which lines are likely to be delayed at what times of day. Genuinely useful for South London commuters and a good excuse to do some proper data analysis.

Started In order to make predictions you need a lot of data. tfl-delay-collector was created to collect that data. It's been running 48 hours as of 27 June 26 so it will be a long time before stage 2.

A lightweight job queue implementation in Go - a learning project that covers goroutines, channels, persistence and worker pools. Go's concurrency model makes it a natural fit for this kind of problem.

Planned

A Go service that connects to the Tempest WebSocket API and fires notifications when configurable thresholds are crossed - frost risk, storm approach, UV spikes, lightning within range. First real-world Go project built on existing Tempest infrastructure.

Planned

A tool that generates improv scene setups from random parameters - location, relationship, object, game - with suggested structures and exercises. Built with the Claude API. A deliberately fun project from someone with a background in improv comedy. Think of it as computerising Clive Anderson.

Planned

An algorithm that solves Wordle optimally using information theory - picks the best starting word and narrows down candidates after each guess. A clean showcase of algorithmic thinking with a very shareable output.

Planned

GitHub Actions workflows to automatically build and push Docker images to a container registry on every push to main, replacing the current manual deployment process across all tempest repos.

Complete

I've often been aware of things in general that people say can't be done because it's very difficult. Maths has a lot of these so I set out to understand - for myself - why they're impossible. Follow along in beautiful-dead-ends as I try and put the pieces together.

In Progress

It turns out that Artifical Intelligence and Machine Learning are quite important topics. When I joined DeepMind as their first IT guy I didn't realise how important it would be but I took the opportunity to learn as much as I could. In what-i-learned-about-ai I've tried to document all of it in an accessible way. Unfortunately I quickly discovered that explaining the equations leads to explaining the equations that explain the equations and it just got messy. So this repo is essentially a series of cheat-sheets for me but I hope others find it useful.

uk-climate-analysis is my attempt to address a certain aspect of the "warm summers" we keep having. I think it's the only repo where I get angry about something!

sen-exclusions-analysis was inspired by a conversation with my social worker wife. I didn't actually find what I went looking for but I did find something else.

crime-stats-uk - the main thing I learned from this is that the UK's largest police for - the one that covers when I live - is absolutely useless at supplying data to aata.police.uk.

A tool I wrote initially for use at DM. Export all your support tickets and your company knowledge base, run this script on the collected data, Receive data that can be pasted into a Tier 0 LLM powered support chatbot. There's three versions the only difference originating from me testing which mainstream LLM (Gemini, Claude, ChatGPT) returned the best interpretation of the supplied documetation.

ha-blind-spots Do you have Home Assistant? Do you have motion sensors and smart lighting? Does every motion sensor in room X control a light in room X? No? You're missing an automation. That's it - it looks for sensible matches between inputs and outputs (e.g. a light and a motion sensor as opposed to a humidity sensor and a robot vacuum) and lists things that you might have missed.

mlops-homelab The scripts I used to build a basic ML Workflow in my homelab so I can keep my MLOps knowledge up to date and attempt to deploy large models on a Kubernetes cluster running on some Raspbery Pis. Because I can.

content-coding engineering pipeline that implements academic qualitative content analysis paradigms at scale.

Google DeepMind (2013-2025) - IT Operations Manager. Built and scaled the global IT operations framework from early-stage startup through Alphabet acquisition, supporting 6,000+ personnel including 4,000+ AI researchers. Designed infrastructure for landmark AI milestones including AlphaGo, AlphaFold, and Gemini. Engineered the data privacy architecture for DeepMind Health, processing live NHS patient data under medical regulatory frameworks.

Shazam (2009-2013) - IT Operations Manager. Scaled infrastructure during hyper-growth, pioneering predictive auto-scaling and leading the transition to Infrastructure-as-Code with Puppet.

Research In Motion (BlackBerry) (2003-2008) - Provisioning Infrastructure Engineer. Maintained 24/7 operational reliability of RIM's global carrier provisioning infrastructure (WebLogic/Oracle) at peak scale - supporting 85 million active devices across virtually every carrier and MVNO worldwide, underpinning service entitlement, service book delivery, and carrier billing for RIM's entire consumer device estate.

Apple UK (1989-1996) - Technical Support Engineer. Part of Apple's foundational UK expansion.

• Enterprise infrastructure architecture at scale
• Python automation and API integration
• DevOps, IaC, and real-time telemetry
• AI governance, data privacy, and compliance frameworks
• Leading global engineering teams and multi-million pound budgets

• Go - building a WebSocket-based lightning alert service as my first real Go project
• Applying Python skills I've used professionally for years to larger structured projects for the first time
• Exploring the intersection of my DeepMind experience with the socio-legal AI governance frameworks I'm studying at the Open University

I'm actively looking for senior roles in infrastructure, platform engineering, technical operations, or AI governance - particularly in organisations where technical depth and leadership experience both matter.

📧 stevet@me.com | 📍 South London

This profile is a work in progress - more projects incoming.