Python / golang engineer

Backend Engineer — Python / Go What you'll do NATS JetStream pipelines — durable consumers, KV buckets, mTLS, ack policies that survive a flaky link. Sharded per-tenant workers handling thousands of msgs/sec without head-of-line blocking. In-memory indices (ring buffers, sorted maps) for sub-ms seek-by-time across the fleet. FastAPI on asyncpg + PostgreSQL — PostGIS, ltree, advisory locks, recursive CTEs. Go for the hot paths Python's GIL can't hold. Own the deploy: Dockerfile, compose, systemd, rollout. One-command install on a fresh box. "Works on my laptop" doesn't count. What we want 5+ years production backend. Idiomatic Python and Go. asyncio at scale — scars and pattern instincts. Message broker depth (NATS JetStream a plus): redelivery, ordering, dedup, partitioning. PostgreSQL beyond CRUD — recursive CTEs, advisory locks, pool deadlocks you've actually debugged. mTLS, JWT, Envoy aren't black boxes. You ship the whole thing — code, container, config, health checks, runbook. You read flamegraphs, tcpdumps, consumer state, and form opinions that hold up. Packaging code into deb / snap form factor and having designed auto updates for such packages. Bonus — robotics/drones/aerospace, PX4/MAVLink, GStreamer/WebRTC, PostGIS at fleet scale. A track record of rewriting bloated Python into tight Go. How we work — Small team. Real customers. Branch to production inside a week. If it doesn't work in the field, it doesn't count. Hard truth If you've been shipping mostly with AI agents and can't fully explain what comes out of them, this isn't the role. We use AI heavily — as a force multiplier, not a substitute. The bar is: read your own diff, defend every line, explain why the consumer config you just changed won't blow up under load. Interviews walk your code without AI. We'll know fast. "It worked when the agent wrote it" isn't a debugging strategy. When telemetry stops flowing in production, nobody is prompting their way out. "Deploy is the SRE team's problem" is not a sentence used here. What runs in production is your responsibility, not the model's. If that energizes you, apply.