Ajay Akilan

Owned the move of the messaging services from virtual machines and configuration management onto Kubernetes, building the infrastructure definitions and deployment pipelines with verification embedded rather than added afterwards. Deploys went from hours to under three minutes, rollbacks to under two.

Led the replatform of the public edge layer across the production fleet on two clouds — phased traffic cutovers, automated verification and failover validation at each stage, no customer impact, edge cost on the largest cluster halved.

Built the automated release-verification framework for the messaging services. A full multi-cluster release went from about seven hours to fifty-two minutes, with verification running against every pod instead of one sample per cluster.

Ran the active-active rollout for the chat backend and edge services across fourteen of fifteen production clusters, owning the networking, scheduling and TLS lifecycle work underneath it.

Wrote the Python CLI that all traffic cutovers run through: resolution tracing with live health data, bulk weight changes behind a preview–confirm–apply flow, an audit line per change. Adopted by DevOps teams elsewhere in the company.

Built production clusters from nothing on three clouds, including one region delivered end to end — infrastructure, five services, active-active enablement, edge migration — and on segregated accounts meeting government data-residency requirements.

Built out observability for the services: hundreds of dashboards, severity-tiered alerting tied to runbooks, and service level objectives defined with the development team rather than handed to them.

Led cost optimisation for the services I own — right-sizing, storage and capacity-mode tuning, cache sizing, tightening requests and limits against observed usage — for per-cluster reductions between 29 and 77 percent and six-figure annual savings.

Incident commander for the messaging estate through hundreds of P0 and P1 events; on call since my third month and the standing escalation point across India, US and China. In five years, no customer-impacting incident caused by a change of mine.