IPv6 dual-stack rollout
Dual-stack networking for ~700,000 public databases in 30+ AWS regions, on a split-horizon DNS architecture that resolves correctly from inside and outside the VPC.
I work on the networking layer underneath AWS's managed databases: DNS architecture, IPv6 at fleet scale, and the availability machinery that keeps roughly 700,000 public databases answering queries. Lately that also means pointing AI at operations, like a retrieval tool for on-call engineers that cut incident investigation time by 74%.
I grew up in Gaighat, in southeastern Nepal, came to Texas for college, and stayed. The longer story is on the about page; the resume has the full record on one page.
$ dig rabindrayadav.com ANY +noall +answer rabindrayadav.com. 3600 IN TXT "Rabindra Yadav, SDE II, AWS RDS networking" rabindrayadav.com. 3600 IN LOC 30 16 N 97 44 W rabindrayadav.com. 3600 IN A 192.0.2.53 rabindrayadav.com. 3600 IN AAAA 2001:db8::35 rabindrayadav.com. 86400 IN MX 10 gmail-smtp-in.l.google.com. ;; Query time: 4 msec
| 700,000+ | public databases migrated to dual-stack IPv6 across 30+ AWS regions | 2025-2026 |
| 99.99% | DNS availability restored and held after a fleet-wide API throttling event | 2025 |
| 74% | less time spent investigating incidents, with a RAG tool built on Bedrock | 2025 |
| 200+ | enterprise adoptions of IPv6 in the first 60 days after launch | 2025 |
| $3M/yr | infrastructure saved by moving Lambda's networking into its own gRPC service | 2024 |
| 180 s | faster Lambda cold starts from parallel initialization | 2024 |
Dual-stack networking for ~700,000 public databases in 30+ AWS regions, on a split-horizon DNS architecture that resolves correctly from inside and outside the VPC.
Retrieval over 10,000+ internal technical documents on Amazon Bedrock. Ask about a failure mode, get the relevant runbook sections back in seconds.
Lambda's networking logic moved into a dedicated gRPC service with parallel initialization. $3M a year saved, cold boots 180 seconds faster.
Shrinking the DNS window customers wait on when an RDS database fails over, projected around 70% faster across millions of Multi-AZ databases.
As of July 2026
The best way to reach me is email: rabindrayadav9596@gmail.com. I usually reply within a day. I'm also on GitHub and LinkedIn.