Embedding VC
AI Engineer
Palo Alto, CA · Hybrid · Full-time · Posted Oct 22, 2026
- Experience
- 3+ years of hands-on ML engineering, or strong research experience with production-quality code.
About the job
Trains, fine-tunes, and serves generative models for interactive AI avatars, covering data, model development, and low-latency realtime inference.
What you'll do
- Train generative models for avatar video, facial animation, and talking-head synthesis; curate and manage multimodal training data.
- Optimize models for realtime serving and build training, evaluation, and deployment infrastructure on GPU clusters.
- Track model quality and collaborate with backend, product, and research teams to ship and improve avatar features.
What they ask for
- 3+ years of hands-on ML engineering, or strong research experience with production-quality code, and a record of shipping deep learning models.
- Deep expertise in PyTorch and modern deep-learning workflows; ability to read, reproduce, and extend research papers.
- Strong background in at least one listed area, including diffusion models, video generation, computer vision, or multimodal AI; solid Python engineering and distributed multi-GPU or multi-node training experience.
- Proven realtime inference optimization experience, including technologies such as TensorRT, ONNX, Triton Inference Server, or CUDA.
Nice to have
- Experience with talking-head or audio-driven animation, lip-sync, digital humans, or related avatar technologies.
- Publications at top venues or notable open-source contributions.
Tools
PyTorch, DeepSpeed, FSDP, Ray, Kubernetes, TensorRT, ONNX, Triton Inference Server
AI focus
Generative AI for realtime interactive avatars, including diffusion, video, multimodal, speech, and neural-rendering models; inference optimization for live streaming.
Model-written summary · the original posting is authoritative