UNITH

• 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 • 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) • Implement new visual features (expressiveness, movement, lip-sync improvements) • Support avatar customization capabilities • Production research enhancements into the real-time video pipeline