Job description

Role Summary: We are seeking an ML Research Scientist to develop and critically evaluate multimodal and decision-making models that can move from research experiments into working systems. This role is for someone who combines research depth with the ability to build, test, and critically evaluate working models. Location: Miami, FL (or willing to relocate) What You'll Do: • Train and evaluate models using combinations of language, images or video, structured state, and temporal sequences. • Analyze datasets, identify quality issues, visualize distributions, and detect shifts between training and deployment data. • Modify model architectures to support new inputs, outputs, objectives, and internal diagnostics. • Design controlled experiments and evaluation methods that distinguish meaningful improvements from misleading metrics. • Move promising research from offline experiments toward simulation, deployed inference, and real-world validation. Must-Haves: • Strong machine-learning fundamentals. • Experience with deep learning, multimodal models, representation learning, imitation learning, reinforcement learning, generative models, or sequential decision-making. Nice-to-Haves: • Robotics or embodied-AI experience is highly desirable (particularly work with robot policies, vision-language-action models, OpenPI, OpenVLA, LeRobot, manipulation datasets, simulation, or physical hardware). Note: This is a preferred differentiator rather than a strict requirement. • Experience with PyTorch, model fine-tuning, dataset tooling, experiment tracking, and production inference.

Requirements

Must-haves

Location: Miami, FL (or willing to relocate)
Train and evaluate models using combinations of language, images or video, structured state, and temporal sequences.
Analyze datasets, identify quality issues, visualize distributions, and detect shifts between training and deployment data.
Modify model architectures to support new inputs, outputs, objectives, and internal diagnostics.
Design controlled experiments and evaluation methods that distinguish meaningful improvements from misleading metrics.
Move promising research from offline experiments toward simulation, deployed inference, and real-world validation.
Strong machine-learning fundamentals.
Experience with deep learning, multimodal models, representation learning, imitation learning, reinforcement learning, generative models, or sequential decision-making.

Nice-to-haves

Experience with PyTorch, model fine-tuning, dataset tooling, experiment tracking, and production inference.

Application prep

Use the role’s language and application details while tailoring your materials.

Keywords to mirror

mlai

Platform: dover
Effort: medium

Benefits

Benefits detected in the listing.

Vision