Project Title Senior Backend AI Engineer – Production AI Pipeline (Execution-Only, 4 Weeks) ⸻ Project Description We are hiring a senior backend AI engineer to execute, stabilize, and harden an existing AI media pipeline for a live consumer application. This is not a research role. This is execution, reliability, and delivery. The system already exists in partial form. All specifications are defined and will be provided. The engagement is strictly time-boxed to 4 weeks. ⸻ Scope of Work (Execution-Only) You will work on an existing backend AI system with the goal of making it production-ready and stable. Backend & Infrastructure • Python backend (FastAPI or equivalent async framework) • Backend-owned AI execution (no client-side generation logic) • Dockerized services • S3-compatible object storage • Redis-based job queues, state, and locking (to be hardened) AI / Media Pipeline • Consolidate and stabilize an existing SDXL + ControlNet + identity-conditioning pipeline • Implement production-grade identity locking (single identity lock reused downstream) • Ensure deterministic, repeatable outputs • No model training • No experimentation • No research Orchestration & Reliability • Async job orchestration • Retry logic and failure handling • Job persistence and idempotency • GPU worker lifecycle management GPU / Ops • GPU workers on RunPod or equivalent • Environment tuning and production hardening • Cost-aware inference execution ⸻ Important Constraints (Read Carefully) • Execution-only (no R&D, no discovery, no model training) • Strict 4-week delivery window • Identity locking is mandatory • Deterministic execution is required • All logic and specifications are predefined and supplied • This is not prompt design or creative work If you prefer experimentation or research, this project is not a fit. ⸻ Required Experience (Non-Negotiable) Please apply only if you have shipped production AI systems. You must have experience with: • Python backend systems (FastAPI / async) • Docker & containerized services • Redis (queues, state management, locking) • GPU inference in production • SDXL in production (not notebooks) • ControlNet + identity conditioning (IP-Adapter, InstantID, or equivalent) • Running AI pipelines on RunPod, AWS GPU, GCP GPU, or similar ⸻ Strong Preference If You Have • Experience stabilizing SDXL pipelines under real user load • Experience with image-to-video or media generation workflows • Ability to clearly state what can and cannot be delivered in 4 weeks ⸻ How to Apply Please include: 1. Brief examples of production AI systems you’ve shipped 2. Your experience with SDXL / ControlNet / identity conditioning 3. Explicit confirmation that you are comfortable with: • Execution-only scope • No R&D • 4-week delivery window Applications proposing research, experimentation, or extended timelines will not be considered. ⸻ Engagement Details • Start: Immediate • Duration: 4 weeks • Commitment: Full-time or near full-time preferred • Budget: Competitive, based on seniority and experience