Sanjay.

// Experience

Where I've worked.

1. Ongoing AWS DevOps Engineer Compunet Connections
   Retrex · AWS staging environment in Sydney · Sep — Oct 2026

   • Built a production-mirroring staging environment in AWS Sydney for a three-service property platform (React frontend, PHP/Lumen API with 421 routes, Go service): VPC, EC2, RDS MySQL 8.4, a private S3 bucket behind CloudFront with Origin Access Control, ACM and Secrets Manager.
   • Kept it at about $68 a month, a third of the $186 production design, by leaving out the NAT Gateway, load balancer and WAF and letting the server's HTTPS port accept traffic only from CloudFront.
   • Moved the 142-table MySQL database into encrypted, private RDS, re-dumping it with hex-encoded binaries and utf8mb4 after a UTF-16 dump corrupted the geometry columns.
   • Secured it end to end: HTTPS all the way to the server, SSM Session Manager instead of SSH, IMDSv2, an instance role that can read only one secret, and a config audit before go-live. A code review of the Go service found 12 deployment issues, including bearer tokens written to request logs.
   • Fixed the launch blockers: .env and .git downloadable from the web root, a 403 on /api/, a 502 behind CloudFront, and every Go signup failing on migrations missing from the repository.
   • Set up monitoring and email: 7 CloudWatch alarms across two regions with SNS alerts, custom memory and disk metrics, and Amazon SES with DKIM, SPF and DMARC all passing.
   • Confirmed the API returns byte-identical responses to the source environment, and wrote the staging and production plans, 29-step build runbooks, a CloudFormation import plan and a CodePipeline / CodeBuild CI/CD plan.
   CMS Marketing · PHP 5 → 8.3 and MySQL 5.1 → 8.0 migration · Jul 2026 — present

   Role: Backend Engineer & Server Administrator · Migration lead

   • Led the migration of a decade-old retail price-intelligence platform (a 15,000-line PHP reporting engine over summary tables with millions of rows) from PHP 5 to 8.3 and MySQL 5.1 to 8.0, through to production cutover on 10–11 Aug 2026.
   • Treated the legacy system as the specification: generated every report from both versions and compared the Excel files cell by cell in Python (openpyxl), so a fix only counted when the diff was zero. Diagnosed 56 defects, fixed and verified 29, with 10,483 products showing zero difference.
   • Tripled concurrent report throughput (an 8-report batch went from 16:08 to 5:06, the worst report from 14:19 to 0:56) by showing that the "slow server" was PHP's session file lock queueing requests, and releasing it with session_write_close() in 20 endpoints.
   • Recovered 4,867 missing prices caused by MySQL 8 discarding ORDER BY inside derived tables; rewrote the "latest crawl per day" query as a deterministic MAX(id) join and fixed 11 more copies of the pattern.
   • Stopped a nightly job from wiping retailers whose crawl had failed (0 of 12 kept their links), replacing its TRUNCATE-and-rebuild with per-retailer updates.
   • Ran the production servers (Ubuntu, Nginx, Postfix, cron on Azure) and the cutover itself, an Azure public-IP swap, then fixed the fallout: a broken chart service, stale /etc/hosts entries and a subdomain-takeover risk. Cut email PTR rejections from 3.6% to 0.9% and deferrals from +4,761 to +19.
   • Traced server-side failures with no code cause: a crawler cron job breaking on a relative include path, and a crontab uploaded from Windows with group-write permission that made cron silently drop every job in it.
   • Added an MD5 deploy gate that caught direct edits on the live server, and kept a 56-incident engineering log recording the root cause, fix and verification for every defect.
   AllenTrans · Docker & Jenkins CI/CD for a fleet management platform

   Role: DevOps Engineer

   • Set up and ran three isolated environments (staging, UAT and production) on the client's cloud VPS for an employee transportation and fleet management platform (Python / FastAPI, PostgreSQL), each with its own database, ports and configuration.
   • Migrated the app from systemd services to Docker, giving every environment the same reproducible runtime.
   • Replaced manual SSH-and-copy deploys with Jenkins pipelines (Git push → Docker build → database migration → deploy) across all three environments, cutting release time by up to 85%.
   • Configured Nginx as a reverse proxy with path-based routing for the backend API and the frontend SPA, and production HTTPS with Let's Encrypt and Certbot.
   • Administered the Linux servers: runtimes, dependencies, permissions and deployment troubleshooting.