Required Senior Backend Engineer - Founding Team
About the role
Own the real-time backend of an AI sales coaching platform that's live, growing 2 week-over-week, and one month into go-to-market. You'll be the second backend engineer in a team of five, working directly with the head of engineering and CEO - building the systems that decide whether the product works: live audio pipelines, LLM orchestration under millisecond budgets, and the engine that turns real sales calls into personalized AI training.
We just found go-to-market fit. Now the backend has to keep up with it. That's where you come in.
A word of honesty:
This is an early-stage role at a small startup in its hardest, most exciting phase, and we'd rather you know exactly what that means before you apply.
You will own large parts of the backend, and ownership here is literal: you'll design it, build it, ship it, and get woken up by it. When a customer's live call drops audio, you'll trace it through the WebSocket, the event loop, and the LLM pipeline yourself. Some of our hardest problems aren't exotic - they're the unglamorous ones every fast-moving team accumulates: a background job that fails without telling anyone, a coaching signal that lands a beat too late, a regression that a replay test should have caught. We know exactly where it hurts, and we're hiring the person who ends it - not the person who files a ticket about it.
If you spent the last few years at a company like Wix or Monday, you already know what excellent engineering looks like at scale - strong typing, clean async code, real observability. What you may not have had recently is the feeling that your work visibly moves the company every single week. Here it does. The trade is real: less process and certainty, far more impact and speed. If that trade excites you rather than worries you, keep reading.
Requirements: Who you are
5+ years of backend engineering with deep Python expertise - and you can point to real-time systems in production that prove it.
Mastery of AsyncIO: you understand the event loop inside out, you know what blocks it and what starves it, and you've kept hundreds of concurrent WebSocket or streaming connections honest - including reconnection, backpressure, and the failure modes that only show up under load.
You've built distributed async pipelines - queues, workers, long-running jobs - and you design them with status, retries, and idempotency from day one, because you've seen what a silently failed job does to user-facing data.
You treat observability as part of the build, not an afterthought: tracing, SLOs, and alerting are how you ship, and "we found it before the customer did" is your default bar.
You write code that survives contact with the outside world: third-party APIs that change under you, OAuth flows that expire at the worst moment, and client versions in the wild you can't force-update. Backward compatibility is a habit, not a chore.
Sane data modeling in PostgreSQL and strict contracts (Pydantic, typing) - because in this product, a wrong number shown to a paying customer is a critical bug, not a display issue.
You finish things. Not "moved to next sprint" - closed, tested, monitored, documented. When a ticket is vague, you sharpen it yourself instead of waiting for a spec.
You're hungry. Maybe you've spent years at a large Israeli tech company and want your work to count for more. Maybe you've always wanted a founding-stage seat. Either way, you bring energy, not just experience - you ship without being pushed, follow a bug across service boundaries without being asked, and treat the product like it's yours. Because it will be.
Strong plus
Real-time voice AI - live STT (Deepgram or similar), conversational audio pipelines (Pipecat, WebRTC), or LLM realtime APIs. A meaningful part of this role is making an AI voice simulation engine reliable.
Nice to have
RabbitMQ/FastStream.
WebSocket scaling patterns (Redis Pub/Sub).
This position is open to all candidates.