■ Profile
AI Platform Engineer
I build the platforms, infrastructure and identity systems that let a small team ship like a large one, and I hold the AI agents writing that team's code to the same standard as people.
I design the architecture and run the infrastructure, identity and security that other engineers build on, and I write the engineering standards they work to.
Much of my recent work is about making AI-assisted delivery trustworthy. Agent-written code goes through the same checks as anyone's: a written spec before code, tests first, a hook that stops an agent finishing until typecheck, lint and tests pass, and adversarial review by a model from a second vendor.
I do my best work in small teams. I set technical direction by writing architecture decision records, choose platforms on measured evidence, and automate any job nobody should be doing by hand.
A Claude Code platform for a whole engineering team, built around a golden path from spec to merge. It is a plugin of 21 skills and 3 hook events for spec-first authoring, test-driven implementation and cross-vendor adversarial review, plus a versioned engineering-standards baseline pushed automatically to about 26 repositories.
Coding agents take specified work, build in sandboxed containers with per-run credentials, pass rubric-graded review gates and merge on automated criteria. It is still being built, on the same spec-to-merge path. I chose the platform after a spike measured a 3.3 s cold start and a 32 ms warm one.
did:web identity with WebAuthn passkeys and W3C Verifiable Credentials, built from scratch. It is in production as the sign-in for a suite of internal applications and federates to self-hosted DIDs, this domain among them.
Cloud and SaaS infrastructure across GitHub, Cloudflare, Kafka, Databricks and observability, run through Terraform Cloud with GitOps: plan on pull request, apply on merge, daily drift detection. AWS and Microsoft 365 are code too, and joiners and leavers are automated.
A meeting-transcription pipeline on an NVIDIA DGX Spark (NeMo ASR, diarisation and voiceprint speaker naming) for audio that cannot leave the building.
An adaptive maths learning platform, started after seeing in the classroom how far edtech falls short of what teachers need.
An agent that emails forecasts to a sailing yacht at sea.
I studied Electrical and Electronic Engineering (BEng), then trained as a maths teacher (PGCE, UCL) and taught secondary mathematics for two years. In 2021 I taught myself full-stack development and moved into software, and I have worked in platform engineering since 2023.
Two years in a classroom left me judging tooling by whether people pick it up and use it well.
Patrick Vieira is a platform engineer based in London — not the footballer of the same name. He designs and runs the foundations other engineers build on: architecture, infrastructure, identity and security.
Platform engineering, AI-assisted software delivery, infrastructure as code and passkey-based identity, mainly with TypeScript, Python, Cloudflare and Terraform.
An internal developer platform for AI-assisted delivery, an agent execution and evaluation platform, a passkey and verifiable-credential identity system, infrastructure as code across cloud and SaaS, and an on-premises speech pipeline.
He holds it to the same bar as anyone's code. A spec comes before any code and tests come first; a hook won't let an agent finish until typecheck, lint and tests pass; and a model from a second vendor reviews the change.
By email at hello@patrickjv.com, or through LinkedIn or GitHub.