Senior Backend Engineer, Generation Platform

Overview

EhudAI is an AI-native platform for filmmakers and storytellers. Behind the product sits a distributed generation platform that fans creative work out to a range of AI providers for image, video, and language models — jobs that run from seconds to many minutes, cost real money the moment they start, and have to be accounted for precisely.

This is a distributed systems role, not a machine learning one. We don't train models and this isn't a prompt engineering job. If your instinct on reading "AI platform" is to think about fine-tuning, it isn't the right fit. If your instinct is to ask what happens when an upstream job is accepted, billed, and never acknowledged — read on.

What you'll be working on

• Owning our generation platform: the services that submit, track, and settle every creative job users run
• Designing a clean provider abstraction so that adding a new model or vendor is a configuration change rather than a new integration
• Building the reliability layer around inherently unreliable upstream APIs — retries, timeouts, idempotency, reconciliation, and honest progress reporting
• Owning the credit and metering system that connects user activity to provider cost, where correctness is non-negotiable in both directions
• Shaping our asynchronous worker topology: queue design, concurrency, scheduling, and graceful shutdown
• Deepening our observability so we can see what the pipeline is doing at every stage and catch problems early

Who we're looking for

• 5+ years in backend engineering, with strong Python
• Deep experience with asynchronous job systems — Celery, RQ, Sidekiq, Temporal, SQS consumers, or similar. You should have real opinions about idempotency, at-least-once delivery, and what happens when a worker dies mid-task
• Strong relational database skills: transactions, isolation levels, and constraint design. You can reason about what two concurrent workers do to the same row
• Experience integrating unreliable third-party APIs at scale — backoff, partial failure, reconciliation, and cost tracking
• Experience with money-adjacent code, or a clear instinct for why it demands a different standard
• You write things down, and you design for the failure case first

Nice to have: familiarity with AI provider APIs, Flask and SQLAlchemy, AWS, and observability work.

You do not need an ML or data science background, and you don't need front-end skills — though you'll occasionally follow a problem across the boundary.

Our tech stack

• Backend: Python, Flask, SQLAlchemy, Celery
• Data: PostgreSQL, Redis
• AI: LangChain, LangGraph, Anthropic Claude, OpenAI, and a range of image and video model providers
• Infrastructure: AWS, Docker, Terraform, GitHub Actions
• Frontend: TypeScript, React, Next.js
• Payments: Stripe, Chapa

You won't have touched all of it. We care that you're strong on distributed backend fundamentals and curious about the rest.

Why should you work at EhudAI?

• You'd own the most important system in the company outright, with real autonomy over its design
• The problems are legitimately hard: distributed state, unreliable upstreams, and accounting that has to be exactly right
• Small, senior team — your work ships in weeks and you're never more than one conversation from a decision
• The domain is genuinely interesting. Creative tooling for filmmakers is a more satisfying thing to build than most backend work
• Meaningful early-stage equity alongside a competitive salary
• A team that treats engineering judgement, not volume, as the thing worth hiring for

How to apply

Write to hello@ehudai.com. Tell us about a time an asynchronous system you owned lost or double-processed work, and what you changed so it couldn't happen again.