Machine Learn Engineer, Video Generation

Description

Summary: Own and scale real-time video synthesis for lifelike digital humans, bridging AI research and production systems in a hands-on, production-focused ML role. Highlights: 1. Own and scale real-time video synthesis for lifelike digital humans 2. Production-first ML role bridging research and deployment 3. Focus on latency, quality, and reliability at scale Machine Learn Engineer, Video Generation**Hybrid · Tech Team · Full\-time** Barcelona, Spain In a few words* Own and scale **real\-time video synthesis** for lifelike digital humans * Production\-first ML role bridging **research** **deployment** * Focus on **latency, quality, and reliability** at scale * Barcelona (hybrid) or remote in Europe \| €45k–€55k **Why this role is exciting:** You’ll work on cutting\-edge digital human technology where your production optimizations have immediate, visible impact on real users and global enterprise customers. About UNITH At **UNITH**, we’re transforming customer journeys with conversational AI. Listed on the ASX, we create **lifelike digital humans** using cutting\-edge synthetic facial movement, voice engineering, and conversational design. Our digital humans speak **60\+ languages** with **600\+ voices**, redefining how businesses interact with customers worldwide. The Role We’re looking for an experienced **Production ML Engineer** to take ownership of our **video synthesis pipeline**. This is a hands\-on, production\-focused role where you’ll bring AI research to life at scale. You’ll work at the intersection of **computer vision, ML infrastructure, and real\-time systems**, ensuring our digital humans run reliably, efficiently, and with ultra\-low latency — without sacrificing visual quality. ️ What You’ll DoProduction Engineering (Core Focus)* Own production video synthesis services and deploy/optimize models for real\-time performance * Reduce inference latency to meet a **\<2\-second target** for streaming conversations * Monitor and improve video quality metrics and debug production issues * Implement model versioning, A/B testing, and safe rollback procedures Integration \& Optimization* Act as the bridge between AI research and production systems * Integrate new models into the existing pipeline * Design video synthesis APIs (gRPC, REST) and work with event\-driven architectures * Optimize GPU utilization, implement caching strategies, and collaborate on service orchestration * Handle TTS integration services (Voice Connectors) Feature Development* Implement new visual features (expressiveness, movement, lip\-sync improvements) * Support avatar customization capabilities * Production research enhancements into the real\-time video pipeline Tech Stack* Python, PyTorch, AWS, Docker, Kubernetes, GPU instances * gRPC services for streaming synthesis * S3, Redis, RabbitMQ ✅ What We’re Looking ForMust\-Have* 3–5 years of experience deploying ML/CV models to **production** (not just training) * Strong hands\-on experience with **PyTorch or TensorFlow** * Practical optimization experience (quantization, pruning, model serving, GPU resource management) * Experience with video generation or real\-time video processing and latency/quality trade\-offs * Strong Python skills for backend services (FastAPI, Flask) and ML serving (TorchServe, ONNX Runtime) * Production infrastructure experience (Docker, AWS, CI/CD pipelines) * Strong debugging skills and ability to collaborate across research and backend teams Nice\-to\-Have* Experience with audio\-driven avatars, face reenactment, GANs, Diffusion Models, or NeRFs * gRPC, RabbitMQ, Go, or video streaming protocols (HLS, WebRTC) * Publications or open\-source contributions in computer vision What Success Looks Like**First 6 months** \- Ownership of core synthesis services with improved reliability and monitoring * Successful deployment of at least one research model into production * Measurable improvements in latency or video quality **First 12 months** \- Streaming video delivery with **30–50% latency reduction** * Production rollout of visual improvements (expressiveness, movement) * Recognized as the go\-to production ML expert bridging research and deployment What We OfferCompensation \& Flexibility Salary: **€45,000 – €55,000**, depending on experience Hybrid work in Barcelona or remote options within Europe Impact \& Growth* End\-to\-end ownership of critical production systems * Unique role bridging cutting\-edge CV research and real\-world deployment * Challenging problems in real\-time ML, latency optimization, and scalability * High\-impact work in a **small, senior team (12 people)** * Opportunity to shape ML infrastructure as the company scales Additional Perks ️ Office in the center of Barcelona Work from anywhere ️ Lunch compensation when in the office Private health insurance with Alan Travel allowance (for team members living 10km\+ from the office) Flexible benefits (tax\-free under Spanish legislation) ️ ClassPass discount How to Apply: Submit: * Your CV highlighting **ML production experience** * A short motivation (3–5 sentences) covering: \- Your experience deploying CV/video models to production * One project where you reduced inference latency * Why digital humans excite you Apply via the **Easy Apply** button, or reach out directly to **joyce@unith.ai** — creativity is welcome Recruitment Process 1\. Technical interview with the tech team (60 min) 2\. Small take\-home assignment 3\. Assignment review (60 min) 4\. Reference check 5\. Call with Joyce (30 min) * ️ **Timeline:** 2–3 weeks from application to offer **Ready to make digital humans faster, better, and more reliable?** Apply now