Autonomy Software Engineering Internship, World Modeling

Persona AI is developing and commercializing rugged, multi-purpose humanoid robots that perform real work. They are seeking an Autonomy Software Engineering Intern to help build the world modeling layer of their autonomy stack, working hands-on with real-world industrial video data and fine-tuning models for humanoid robots.

What you'll do:

• Design and implement autonomy and behavior coordination algorithms for an industrial humanoid platform
• Work with manipulation, locomotion, and perception engineers to integrate robot skills into autonomous robot behaviors
• Implement data curation pipelines that enable our robots to understand a variety of real-world industrial tasks
• Develop synthetic data generation pipelines for post-training of humanoid world models
• Design and implement pipelines for autonomous behavior evaluation, benchmarking, and regression testing
• Turn raw industrial video footage (shipyards, steel fabrication, welding cells) into structured, high-quality training data
• Design and evaluate reasoning benchmarks that measure how well a fine-tuned world model grounds language queries in real shipyard and fabrication-plant scenes

What they're looking for:

• Currently pursuing a BS, MS, or PhD in Robotics, Computer Science, Machine Learning, Electrical Engineering, or a related field; recent graduates are also welcome
• Hands-on experience training, fine-tuning, or evaluating modern deep learning models in PyTorch
• Strong Python skills; comfortable writing data processing scripts, training loops, and evaluation harnesses
• Working familiarity with at least one vision-language model, large language model, or video understanding model
• Practical experience building data pipelines: ingestion, filtering, labeling workflows, deduplication, train/val/test splits
• Comfortable working in Linux environments with Git, containers (Docker), and cloud GPU infrastructure
• Self-directed and able to make steady progress on an open-ended research problem with weekly check-ins rather than daily direction
• Direct experience with NVIDIA Cosmos or comparable world foundation models
• Exposure to NVIDIA Isaac Sim, Omniverse, or OpenUSD for synthetic data generation and scene variability
• Familiarity with humanoid or mobile-manipulator robotics, ROS/ROS 2, and the broader autonomy stack
• Experience with motion capture data, egocentric video capture, or first-person dataset construction
• Exposure to behavior trees, task and motion planning, or LLM-driven planning architectures

Benefits:

• Full access to advanced prototyping tools, labs, and the freedom to experiment and innovate
• Flexible work environment
• Meaningful technical mentorship