Machine Learning Researcher

Founding ML Research Scientist

Seattle, WA (On-Site)

Our client is a stealth-stage AI startup developing a real-time, human foundation model that brings social and emotional intelligence to voice, face, and body for next-generation interactive experiences. The team is well-funded, early-stage, and focused on building core foundational models rather than application-layer features.

This role leads the end-to-end training of large multimodal, autoregressive models that jointly reason over text, speech, facial expression, and body language in real time. You will own research, data strategy, and large-scale training pipelines to power lifelike interactive avatars that respond with nuanced expressions, gestures, and tone frame by frame.

Responsibilities

• Design and train large multimodal autoregressive models across text, audio, and video (face and body) for real-time interaction.
• Develop model architectures, objectives, and optimization strategies to capture fine-grained human signals (e.g., prosody, micro-expressions, body pose dynamics).
• Build scalable training, evaluation, and deployment pipelines for low-latency inference in production environments.
• Define data collection, curation, and labeling strategies for multimodal human interaction datasets, including safety and privacy guardrails.
• Establish rigorous offline and online evaluation frameworks for social/emotional intelligence, realism, and responsiveness.
• Collaborate closely with founding researchers and leadership to translate research breakthroughs into product-ready capabilities.
• Mentor junior researchers/engineers and help set technical standards, coding practices, and research culture in a small, high-ownership team.

Qualifications

• 3 years of experience training large-scale multimodal or language models, autoregressive architectures, or closely related foundation model work (industry or post-PhD).
• Strong background in deep learning for one or more of: speech, audio, computer vision (face/body), or sequence modeling.
• Hands-on experience implementing and training transformer-based or similar architectures with modern ML frameworks (e.g., PyTorch, JAX, or TensorFlow).
• Proven track record of end-to-end model development: problem formulation, experimentation, training, evaluation, and iteration.
• Solid software engineering skills, including writing production-quality Python and working with large-scale training infrastructure on GPUs/TPUs.
• Comfort working 5 days per week onsite in the Seattle area in a fast-paced, highly collaborative environment.

Preferred Skills

• PhD in Computer Science, Electrical Engineering, Robotics, or related field with research in multimodal learning, generative models, or human–computer interaction.
• Experience building or training MLLMs, conversational agents, or avatar/embodiment systems that combine vision, audio, and language.
• Publications at top-tier ML or vision/graphics conferences (e.g., NeurIPS, ICML, ICLR, CVPR, ICCV, SIGGRAPH).
• Prior experience in early-stage startups or small, fast-moving research teams with high ownership.
• Familiarity with real-time inference optimization (quantization, distillation, on-device deployment) and streaming architectures.