An engineer's deep-dive into the platform architecture described by Vasilios Syrakis in his video "I was laid off by Atlassian." Reconstructed from his own walkthrough — no internal code, data, or users; systems only.
📺 Source: "I was laid off by Atlassian" by Vasilios Syrakis · ~38 min · channel @vsyrakis. Vasilios was a senior systems engineer at Atlassian for ~8 years, laid off in March 2026 in a ~10% workforce cut. This document credits him fully and exists to make his (deliberately public) knowledge-share easier to study. All architecture below is his account.
⚡ TL;DR
Over eight years, Vasilios built Atlassian's self-service edge load-balancing platform — the layer that puts Jira, Confluence, Bitbucket and Statuspage on the internet. Five systems, stacked:
- Open Service Broker (OSB) — a self-service API so dev teams provision load balancing without filing a ticket.
- Envoy proxy — the open-source, cloud-native replacement for expensive enterprise load balancers.
- Sovereign — the Envoy xDS control plane he built and open-sourced; turns dynamic data into live proxy config.
- AWS infrastructure — ~2,000 Envoy proxies across ~13 regions, provisioned as code via CloudFormation.
- The AMI pipeline — Packer + SaltStack baking a hardened, observable, sidecar-loaded proxy image.
On top of that he centralized auth, authz, rate limiting, DDoS protection and access logging at the edge — solving them once instead of across a thousand backend services. The back half of the talk is the part engineers skip at their peril: maintenance, code churn, diplomacy, and mentoring.
🧭 The interview that defined the job
The hiring loop is worth a beat, because the last interview literally scoped his first two years of work.
- Coding quiz — HackerRank, full marks.
- Technical 1 — Handed a Cloudflare white paper on custom domains, left alone ~10 min to read, then quizzed on it plus microservices, architecture and containers.
- Technical 2 — A live troubleshooting exercise: prompt the interviewer for information to diagnose a real past Atlassian incident — an application bug that cascaded into a denial of service. Plus a question on latency-based DNS / Route 53.
- Values interview — He flipped it: "Picture 12 months from now — what would I need to have achieved for hiring me to look like a good decision?" The answer: build a web app for self-service load balancers — internal-developer equivalent of AWS Application Load Balancers. He said he could build it in Python. They believed him. Hired.
💡 Takeaway: Asking the "what does success look like in 12 months" question didn't just land the offer — it surfaced the exact mandate. Know the job before you accept it.