Secure cloud infrastructure, built and run remotely.
I design, build and run production systems on AWS: infrastructure as code, deploy pipelines, containers and real-time media. I handle it end to end, from the first Terraform file to the 3am incident, and I write it all down.
Get in touch See recent workWhat I do
Hands-on engineering, not slide decks. You get working infrastructure, the code behind it, and runbooks your team can follow.
Infrastructure as code
AWS environments in Terraform: networking, EC2 and autoscaling, Aurora, ElastiCache, EFS, S3 and IAM, with every change planned and reviewed before it's applied.
CI/CD pipelines
GitHub Actions with OIDC (no long-lived keys), test and static-analysis gates, and deploys that remove deleted files and check the result afterwards.
Containers and compute
Docker workloads, AWS Batch and Fargate Spot jobs, and Lambda-launched short-lived workers that shut themselves down, so you pay only when there's work.
Security hardening
Least-privilege IAM, secrets held in SSM rather than in code, WAF and rate limiting at the edge, and relay and media services locked down against abuse.
Reliability and incidents
Health checks that test the real thing, CloudWatch alarms that point at the cause, and calm, written-up recovery when something breaks.
Cost control
Measured capacity planning, scheduled scaling, Spot where it's safe, and caps on pay-per-use services before they turn into a surprise bill.
Migrations
Moving domains, DNS, email and hosting between providers without downtime, including the records everyone forgets.
Recent work
A live online-tutoring platform: one-to-one video lessons, recording, and payments to tutors.
Real-time video learning platform on AWS
- Clustered media servers. Scaled a single OpenMeetings/Kurento node to a multi-node cluster and load-tested it with 70 concurrent classes. The tests found and fixed three bugs that only appear once a second node exists.
- Moved video conversion off the media host. Recordings are now converted by AWS Batch on Fargate Spot, at about $0.002 per class, so live lessons don't compete with ffmpeg for CPU.
- Self-terminating worker fleets. Lambda launchers start one EC2 instance per class, and a reaper backstops any that don't shut down. When there are no classes, the fleet costs nothing.
- Hardened the edge. TURN over TLS on 443 for networks that block UDP, a Cloudflare WAF with rate limits, and coturn locked down against relay abuse.
- Disaster recovery that was tested. Database snapshot restores and EFS backup restores were run as real destroy-and-rebuild drills, not just written down.
About
I'm a creative problem solver and explorer, based in London, and I've worked in DevOps for the last few years. It started with programming a Sinclair 48k to make a frog jump. At college and at Goldsmiths, University of London, I brought electronics, hacker culture and 3G into my studies. More recently I've done DevOps for an award-winning UK FinTech.
Away from work I tinker with hardware. Recent projects include reverse-engineering a battery's Bluetooth protocol and wiring a camper van's solar and charging system around a Victron Cerbo.
How I work
Fully remote and comfortable across time zones.
- 01Written first. Plans, decisions and runbooks live in the repo, so nothing depends on my memory.
- 02Verify, don't assume. A deploy isn't done until the live result has been checked, not just the exit code.
- 03Safe changes. Review the plan, keep a rollback ready, and use maintenance windows for anything risky.
- 04Async by default. Clear written updates, with calls when they actually help.
Let's talk
Need a hand with infrastructure, a migration or a pipeline that keeps breaking? Send a short note about what you're working on.
info@collisiondetection.biz LinkedIn