Senior Python/Django DevOps: Production-Ready Async Processing Pipeline

We're looking for a senior Django + DevOps engineer to take a partially-built video-processing module from demo mode to a production-ready deployment. PROJECT CONTEXT Our product is a Django SaaS for sports performance management. One of its modules processes uploaded match videos through a computer-vision pipeline (separate work stream, not part of this job) to produce analytics. The module already has most of its plumbing built — models, API endpoints, async task structure — but needs consolidation and a production-grade deployment split across two platforms: - Web/API layer: Django on Vercel (serverless) - Heavy processing worker: Celery on Railway (persistent container) - Message queue: Redis (Upstash) - Database: PostgreSQL (Neon) - Video storage: Vercel Blob WHAT YOU'LL DO - Finalize Django migrations and audit multi-tenant data isolation (all data must be scoped per client/organization) across views, permissions and query selectors. - Provision and configure the Celery worker + beat scheduler on Railway, and the Redis broker on Upstash. - Propagate environment configuration correctly across Vercel and Railway (they don't share config automatically). - Harden file upload validation and error handling for the async flow. - Add baseline logging/observability per processed job. - Expand automated test coverage for the async task flow. - Critical constraint: heavy dependencies must never load in the Vercel serverless runtime — they must stay isolated inside the worker. WHAT WE'RE LOOKING FOR - Strong hands-on experience with Django in production, including Celery-based async processing. - Experience with serverless deployments (Vercel or similar) AND persistent container platforms (Railway or similar: Render, Fly.io, Heroku). - Comfortable working across Postgres, Redis, and cloud object storage. - Experience auditing multi-tenant SaaS data isolation. - Bonus: experience with video-processing or ML-inference async pipelines. DELIVERABLE A real video upload flows end-to-end through the production worker, correctly scoped by tenant, without breaking the existing Vercel deployment. A short technical brief is available attached to this post.