Backend Engineer — Reliable Async Job Queue for AI API Calls

We're building a backend that makes a lot of calls to OpenAI/Anthropic and a few other third-party APIs, and we're hitting the usual problems: dropped jobs on timeouts, no retry/backoff strategy, rate limits (429s) breaking delivery, and no visibility into what actually ran vs. failed silently. Looking for a backend engineer to help us architect and implement a reliable async job pipeline — jobs get queued, retried with backoff on failure, rate-limited per downstream, and delivered to a webhook with acknowledgment. Comfortable working in Node.js/TypeScript or Python. Scope for this engagement: Review our current setup and recommend an approach (self-hosted queue vs. managed service — open to your recommendation) Implement retry/backoff, rate limiting, and dead-letter handling for failed jobs Set up webhook delivery with basic observability (logs / attempt history) This could grow into ongoing work if it's a good fit. Please include in your proposal: what you've used before for background jobs (Redis+BullMQ, SQS, Celery, RabbitMQ, Cloud Tasks, QStash, Inngest, or similar), and a project where you've handled retries/rate limits against a third-party API.