ML Research Engineer

We’re partnering with a venture-backed AI startup building next-generation visual conversational systems – enabling users to interact with AI through real-time video experiences that feel genuinely human. Their work sits at the intersection of multimodal ML, high-performance infrastructure, and real-time systems. They’re now hiring a Machine Learning Research Engineer to help bring cutting-edge models from research into highly optimized production environments.

This is a hands-on role focused on GPU performance, model acceleration, and scaling multimodal systems in real-world deployments.

Key Responsibilities:

• Collaborate with research teams to productionize experimental models and transition them into reliable, scalable systems.

• Own model performance optimization, profiling inference for latency and throughput improvements using techniques such as quantization, pruning, and distillation.

• Debug and optimize GPU workloads, including CUDA-level performance issues.

• Apply acceleration frameworks (e.g., TensorRT, ONNX, vLLM) to improve multimodal model efficiency across video, speech, and LLM systems.

• Design and implement high-throughput data pipelines for large-scale video and multimedia datasets (petabyte-scale).

• Develop evaluation frameworks to measure model quality and support continuous iteration.

• Work closely with infrastructure teams to build scalable media processing and training workflows.

Key Qualifications:

• 2+ years of full-time experience in ML engineering, ideally in production ML environments.

• Proven experience operationalizing research models into production systems with a focus on inference optimization (latency and throughput improvements).

• Strong proficiency in PyTorch for training and deploying large-scale models.

• Experience running models across large datasets for feature extraction or batch inference workloads.

• Hands-on experience working with GPUs and debugging CUDA-related performance issues.

• Experience with video, audio, or multimodal models preferred.

• Background in a VC-backed startup or high-performing established company is advantageous.

• Willingness to work onsite 5 days per week in Seattle.

• Strong hands-on engineering focus (not a management-leaning or director-level profile).

• Candidates whose ML experience is limited to fine-tuning models without deeper systems or performance work are unlikely to be a fit.