Backend engineer

Backend Engineer — Go, Distributed Systems, NDCRemote · Full-time · WorldwideNorba is the NDC aggregation API built for travel engineering teams. We're looking for an excellent backend engineer to build and operate the Go services that fan out across airline NDC APIs, normalize responses into a canonical schema, and keep the platform reliable, fast, and scalable.What you'll do- Build Go services that run the aggregation engine: concurrent fan-out to multiple NDC connectors, normalization to `models.Offer`, quality scoring, deduplication, and offer merging.- Implement and operate caching, rate limiting, and circuit breakers so an airline outage doesn't cascade and repeated searches resolve in milliseconds from Redis.- Design NDC XML normalization pipelines (versions 17.2, 18.1, 19.2, 21.3) to canonical JSON, including parsing free-text fare rules and structured extraction of penalties.- Build the servicing layer: OrderRetrieve, OrderChange, OrderCancel, OrderReShop against airline systems, with retries, idempotency, and monitoring of every attempt.- Work in a remote, async, written-first team: design in docs, review in PRs, communicate decisions in writing.What we're looking for- Excellence in Go and distributed systems: you've shipped services to production, you reason about concurrency, latency, backpressure, and partial failures.- Ownership: you take a problem end-to-end (from exploring an airline API to having the connector live in production with monitoring) without waiting to be told the next step.- Strong written communication: you document designs, explain trade-offs, and write clear PRs.- Genuine interest in travel technology and airline distribution; you don't need to know NDC yet, but you need to want to.Nice to have- Prior exposure to the travel industry: airlines, GDS, OTAs, or IATA standards (NDC, ONE Order).- Experience at an early-stage startup, where roles are broad and the playbook is yours to write.- Work with complex XML APIs, heterogeneous schema mapping, and data normalization.How to applyIn Norba we read every application.Any application detected as AI-written is rejected immediately; we'd rather read three human lines than three model paragraphs.