| ▲ | sagarw 5 hours ago | |
Location: Pune, India Remote: Yes (IST) Willing to relocate: Yes, Bengaluru; also open to hybrid/onsite there Technologies: Go, PostgreSQL, Redis, AWS (RDS, SQS, S3), Kubernetes, Terraform, Stripe, RabbitMQ/SQS, Prometheus/Grafana, k6. Also Python, TypeScript/React, Elixir. Résumé/CV: https://www.linkedin.com/in/sagar-waidande-1337bb9a/ GitHub: https://github.com/sagarsuperuser Email: sagar10018233@gmail.com Availability: immediately; full-time, or contract/contract-to-hire Backend engineer, 8 years, writing Go since 2019, mostly on systems that move money: real-time trading for 40K concurrent users, payment APIs across 4+ providers at a cross-border payments company, and billing. Since 2025 I've built open-source backend systems full time and load-tested them on AWS, with public, dated benchmarks: * Velox (https://github.com/getvelox/velox), a usage-based billing engine in Go + Postgres + Stripe. Idempotency keys end to end, a unique index that rejects a second invoice for the same period, a reconciler that looks up ambiguous Stripe charges instead of retrying blind, and leader leases with fencing tokens. 12,000-15,000 events/sec of batched ingest on RDS, every event reconciled. Billing leader SIGKILLed at five points mid-run: 0 duplicate or lost invoices. Negative control: drop the index and the same run bills 103 invoices for 40 periods, $2,575 against $1,000. * A notification platform (https://github.com/sagarsuperuser/notif-service) on SQS/RDS/Kubernetes: 2,000 accepts/sec sustained, and a 58-second total provider outage lost 0 of 100,000 messages. Along the way I found a retry bug silently dropping 8,838 of 100,000 messages and proved the fix with an A/B run on AWS. Looking for senior backend roles where correctness matters: payments, billing, ledgers, infrastructure. | ||