Realtime Generative AI Avatar Engineer

LiveX AI Inc.; Palo Alto (CA); On-site; USD 150,000 - 210,000; Full time

Benefits offered by this job

Equity

Benefits package

Career growth opportunities

Job summary

LiveX AI Inc. is building next‑generation realtime, interactive AI avatars that see, listen, and respond with natural speech and expressive motion. We seek a strong ML Engineer to train, fine‑tune, and serve generative models powering these avatars.

You will work across the full model lifecycle—from data curation and training of diffusion, video, and multimodal models, to low‑latency inference optimization for live streaming deployments.

Qualifications

• 3+ years of hands-on ML engineering experience shipping deep learning models.
• Strong expertise in PyTorch and modern ML workflows.
• Experience with diffusion, video generation, or multimodal AI is highly desirable.
• Experience optimizing models for real-time, low-latency inference.

Responsibilities

• Train and fine‑tune state‑of‑the‑art generative models for talking-head synthesis and avatar video.
• Design data pipelines for large-scale video, audio, and multimodal datasets.
• Optimize models for sub-200ms end‑to‑end inference and streaming pipelines.
• Build and maintain training/evaluation/deployment infra on GPU clusters.
• Define and track quality metrics like lip-sync accuracy and latency.
• Collaborate with backend, product, and research teams to ship features to production.
• Stay current with research in generative video and neural rendering.

Skills

PyTorch

Diffusion models

Video generation

Multimodal AI

Streaming inference

Python

Distributed training

Tools

CUDA

TensorRT

ONNX

DeepSpeed

Job description

LiveX AI Inc. is building next‑generation realtime, interactive AI avatars that see, listen, and respond with natural speech and expressive motion. We seek a strong ML Engineer to train, fine‑tune, and serve generative models powering these avatars.

You will work across the full model lifecycle—from data curation and training of diffusion, video, and multimodal models, to low‑latency inference optimization for live streaming deployments.