Feb 2025 – Present · 1 yr 4 mos

• 🛰️ Owned reliability of our in-house UAT provisioning platform — shipped a 3-stage GitHub-outage fallback (probes → automated cutover → locally-running controller), rolled out a local executor across all pools, and added scoped retry + per-pool circuit-breaker logic to keep developer environments available even during upstream incidents.
• 📊 Led the Prometheus → VictoriaMetrics migration end-to-end (org-wide OKR) — designed the prototype, rolled out across non-prod, SaaS, SaaS-blue and production clusters, and migrated every dashboard and alert onto the new ingestion path.
• 🤖 Built an AI-first on-call and DevOps assistant in Slack — multi-tenant engine with channel-to-plugin routing, RCA automation wired into Grafana webhooks, automated PR creation for known-fix patterns, and AI plugins for self-service infra workflows (service scaffolding, region extension, ECS migration).
• 🧠 Extended an internal MCP (Model Context Protocol) server with ClickHouse, Metabase and Kubernetes tooling - gave AI agents structured, audited access to logs, metrics and analytics across all environments.
• 🚀 Productionised two new business services end-to-end — Kubernetes manifests, ArgoCD integration, Datadog observability, egress allow-listing - including a Blocker-priority production launch in a new region.
• 🌐 Consolidated the Helm chart estate into a single OCI-backed monorepo with GitOps adoption and private-endpoint chart pulls for all application charts.
• 🔔 Designed Alert-as-Code on a new monitoring framework with Tofu workflows, smart alert routing, and ClickHouse-based log alerting - migrated multiple team alert stacks onto the IaC pipeline.
• 🛠️ Self-service infrastructure OKR - researched, prototyped and implemented GitOps-enforced creation flows so developers can stand up new services without DevOps in the loop.

Aug 2024 – Feb 2025 · 6 mos

• 🏗️ Migrated the entire Terraform estate from a legacy repo into a consolidated monorepo framework across non-prod, SaaS, production and production-AU accounts (Datadog, notifications, CloudTrail, CodeArtifact, S3, TGW, CloudWatch alarms) and fully deprecated the legacy repo.
• 🔐 Designed and shipped a Vault-backed dynamic-credentials workflow for Redshift with a TTL'd approval gate - integrated via a Slack bot so developers could self-serve time-boxed, audited DB access, replacing the long-lived credential model.
• ⚙️ Improved the shared GitHub Actions framework used by every infra pipeline - added multi-target tf plan/apply support, commit-message-driven plan skipping, shellcheck linting, and retry logic for git operations.
• 📟 Built out the Grafana observability stack (Loki/Mimir/Tempo) - added log-based alerts, Karpenter event alerting, profiling research on ClickHouse, and designed a YAML/JSON dev-input struct so service teams could declare Alertmanager rules without touching infra config.
• 🔁 Enhanced production rollback UX - implemented "Rollback-as-Stack" for tightly-coupled service pairs and a Slack-native flow so on-call could atomically revert linked services.
• 🚀 Stood up CI/CD for a new product service (Highest priority) — first new service I owned end-to-end including pipelines, ArgoCD integration and observability wiring.
• 📦 Upgraded the logging stack (Loki 3.3.2 / chart 6.24.0) and audited & reconfigured every SNS topic and subscription across the AWS estate.
• 🛡️ Moved internal dashboards behind Twingate to remove public attack surface flagged during a security review, and stood up MobSF in SaaS for mobile app security scanning.
• 🤖 Built the first Slack-bot↔infra workflow integration (Redshift user creation) - laid the integration pattern that later automation reused.
• 🚨 On-call from day one - auto-sync chart failure alerts, dashboard fixes for the API gateway, smoke-test pipeline fixes in Jenkins, and infra-wide alert coverage rollout.