Open to opportunities

Hi, I'm Sudhanshu Singh.

DevOps / SRE engineer — I migrated our platform to AWS and now run it solo for 50,000+ users: IaC, CI/CD, observability, and database reliability.

Location
What I run
  • Infrastructure as CodeReproducible, multi-AZ AWS environments
  • CI/CDBlue-green deploys with instant rollback
  • ObservabilityDashboards, alerting, and log search
  • Incident responseLive diagnostics and postmortems
  • Database reliabilityQuery & index tuning, right-sizing

Writing

Notes from production

Deep-dives on the reliability, performance, and cost work behind a 50,000-user platform — the real logs, numbers, and fixes.

View all posts
A latency-versus-working-set curve staying flat while the working set fits in RAM, then falling off a cliff once it spills past the tier's memory limit

2026-08-12 · 6 min read

Right-Sizing a Production MongoDB Cluster: M50 → M40

You can't just click "downgrade" on a production database. Here's the log forensics, index surgery, and aggregation rewrites that shrank the working set enough to take a MongoDB Atlas cluster from M50 to M40 — a full tier down, slow-query time down ~43%, and latency that improved through the cut.

  • mongodb
  • performance
  • cost-optimization
  • sre
  • databases
Read more
Before/after architecture of a zero-downtime cloud migration — a flat, unsegmented setup versus a layered AWS network with private subnets and a private database

2026-08-12 · 8 min read

A Zero-Downtime DigitalOcean → AWS Migration (50k Users)

Moving a live 50,000-user production platform off DigitalOcean and onto AWS without a single customer noticing — rebuilt entirely in Terraform and flipped over with one reversible DNS cutover, with the old stack kept warm as a real rollback net.

  • aws
  • digitalocean
  • migration
  • zero-downtime
  • terraform
  • infrastructure-as-code
  • sre
  • devops
Read more
A delivery status recorded at the moment of API acceptance versus the true outcome — delivered, bounced, or complained — that only arrives later on an event stream

2026-08-11 · 8 min read

The "DELIVERED" That Never Arrived

A notification database was reporting DELIVERED for mail that never arrived. Delivery was recorded the instant SES accepted the API call — not when the message landed — so ~37,000 bounces a day were filed as successes, hiding a steady 52% bounce rate and a bill quietly running at double. The fix was to stop trusting acceptance and wire the real delivery events back on.

  • observability
  • aws-ses
  • email-deliverability
  • metrics
  • sre
  • cost-optimization
  • rabbitmq
  • reliability
  • devops
Read more