Backend engineer · Amsterdam · Available now
I build backend systems other teams depend on, and lately the AI agents that run on top of them. Six years at SAP, most recently on the OpenTelemetry platform that 300+ cloud services report into.
I moved to Amsterdam in November 2026 on a Dutch partner permit, so I have full right to work and need no sponsorship. I'm looking for a backend team here to grow with.
I built the core of SAP Cloud ALM's OpenTelemetry routing platform and the Java and Node.js client library every connected service uses: nearest-datacentre routing, automatic mTLS certificates, and failover across 17 datacentres. Kafka-driven onboarding took new tenants from two days to near-instant, and the HTTP/2 migration I led added 45% throughput.
Onboarding was eating four days of every twelve-day sprint. I built a prototype over a weekend, hardened it, and shipped it to SAP's cloud service teams a week later. How it works.
The SDK quietly turned every "+" in resource attributes into a space, because it decoded them like form data. I replaced the decoder with one that follows the W3C Baggage spec, with tests.
Start from the data. Jira showed where my team's time went: onboarding services, troubleshooting, and fixing onboardings that went wrong took around 31 person-days a quarter. That decided what the agent had to do first.
Earn trust in stages. Service teams onboard their dev and test landscapes through the agent's chat, with help instrumenting our client library. Once they've checked that monitoring works and approved it, the agent pulls all their production tenants and onboards them on its own. People verify where mistakes are cheap; the agent takes over the expensive, repetitive part.
Corrections become fixes. An admin UI captures live conversations. When someone corrects an answer, the correction goes back into the prompt or the knowledge base. Retrieval misses users reported were fixed by tuning chunking, embeddings and a glossary of internal jargon.
Keep it debuggable. Every run is traced with OpenTelemetry and Langfuse, and every config change the agent makes is a commit to version-controlled YAML in GitHub, so it can be audited and reverted. Built on Claude Opus with the Anthropic SDK on SAP AI Core.
M.Tech. (Master of Technology) in Software Engineering from BITS Pilani, studied part-time while working, on a full scholarship. Dissertation on efficient resilience in distributed cloud systems. Before that, a BCA from MIT World Peace University, Pune. I mentored three interns at SAP, and outside work I play guitar.