Research Scientist - Generative Video

📍 Location: Seattle, WA, US - On-site

💰 Salary: $240K - $300K Base + signing bonus

📈 Equity: 0.5% - 1%

🛂 Visa Sponsorship: Available

About Aurora

Aurora

helps

exceptional engineers

discover opportunities at some of the

most ambitious startups worldwide

.

We work closely with companies to identify exceptional talent and match them with roles where they can have real impact.

We are currently helping a

series B startup

in

Seattle

grow their research team.

About the Company The company we are hiring for is building

visual conversational AI

that feels human. Their goal is to power experiences where users can jump on a live video call with AI that responds naturally in real time.

The company is building a human foundation model that operates across text, speech, facial expression, and body language simultaneously. The system aims to understand subtle human signals and respond with lifelike expressions, gestures, tone, and emotion.

Founded in 2024, the company raised

$60M

, has a

team of 14 people

, and is based in Seattle.

The founding team includes researchers and operators from

Apple, Google, Meta, Microsoft, MIT, Oxford, and Carnegie Mellon

, with deep expertise across robotics, graphics, avatars, and machine learning.

What This Role Actually Is

This is best described as a

Generative AI Research Scientist

focused on

video diffusion

, avatar generation, and multimodal real-time AI systems.

It is a highly technical frontier research role centered on training state-of-the-art generative models rather than product engineering.

This is ideal for candidates from top AI labs, research teams, or strong applied research backgrounds in diffusion / video generation.

About the Role The company is hiring a Research Scientist to help build foundational technology in a space where the boundaries are still being defined.

You will work on lifelike, responsive avatars whose facial expressions, gestures, and tone evolve frame-by-frame to generate genuine real-time responses.

The broader mission is to create the first human foundation model capable of understanding and generating signals across language, voice, facial movement, and body language.

This is an opportunity to shape core model architecture and research direction at an early-stage, heavily funded company.

What You’ll Do

• Train and improve diffusion models for image, video, or 3D generation.

• Build real-time avatar systems that generate human-like motion, expression, and emotional responsiveness.

• Work on multimodal modeling across speech, text, vision, and body language.

• Improve quality, latency, realism, temporal consistency, and controllability of generated outputs.

• Collaborate closely with elite researchers and founders on new model architectures and experiments.

• Help define the technical roadmap for human-centered generative AI.

Requirements

• 3+ years of experience training diffusion models for image, video, or 3D generation.

• Strong background in deep learning, generative modeling, PyTorch-scale experimentation, and research iteration.

• Experience shipping or researching state-of-the-art systems in video generation, avatars, graphics, multimodal AI, or adjacent areas is highly valuable.

• Strong ownership, fast execution, and ability to thrive in a small elite in-person team.

Tech Focus

Video Diffusion, Generative AI, Avatars, Multimodal AI, Real-Time Inference, Computer Vision, Deep Learning

Why Join

This is a rare chance to work on one of the most ambitious problems in AI: making AI visually and emotionally human in real time.

The company has exceptional funding, an elite technical bench, and a flat structure where early hires can heavily influence both research direction and company trajectory.

If successful, the technology could power entirely new categories including AI companionship, AI interviewing, AI sales, and human-like enterprise agents.

This is frontier research with real product upside.