Thermo Fisher Scientific
Thermo Fisher Scientific
Thermo Fisher Scientific develops and supplies scientific instruments, reagents, consumables, and software services for organizations working across life sciences, healthcare, and analytical chemistry. Its portfolio supports laboratory research and production, with applications spanning diagnostics, lab workflow automation, and drug discovery. The company’s products and services are used to help research and healthcare teams manage complex laboratory processes and advance scientific development.

Physical AI Research Intern (Remote, United States)

Research and develop physical AI systems spanning computer vision, vision-language models, 3D perception, and robotics. Turn experimental work into reproducible code, deployable services, evaluation reports, and demonstrations.

Description

  • Develop and fine-tune models for detection, tracking, segmentation, pose and activity recognition, and scene understanding
  • Train and assess vision-language models for grounding, dense captioning, temporal question answering, and tool use
  • Create retrieval-augmented and agent-based workflows for perception and action
  • Prototype perception-driven control policies with simulated and real-world data
  • Connect perception systems to planners and task graphs for manipulation, navigation, and safety applications
  • Prepare datasets, define evaluation methods and KPIs, and conduct ablation studies
  • Deploy research systems as dependable services with Kubernetes, Docker, Ray, and FastAPI
  • Implement profiling, telemetry, and continuous integration to support reproducible research
  • Coordinate multi-agent pipelines across perception, reasoning, simulation, and code generation
  • Investigate video understanding, 3D scene grounding, edge inference, multimodal assessment, and sim-to-real policy performance
  • Produce publishable or open-source research, production-ready components, reproducible implementations, evaluation reports, and demos

Requirements

  • Pursuing a master’s or Ph.D. in computer science, electrical engineering, robotics, or a related field, with active publications in computer vision, machine learning, or robotics
  • Advanced Python and PyTorch or JAX skills, including familiarity with CUDA profiling and mixed-precision training
  • Research experience in computer vision plus at least one area such as vision-language models, embodied or physical AI, or 3D perception
  • Ability to progress from published research to implementation, ablation studies, and rigorously tracked results
  • Experience with video models, diffusion, 3D Gaussian Splatting or NeRF, SLAM, or scene reconstruction
  • Previous work involving multimodal grounding or temporal reasoning
  • Familiarity with ROS2, DeepStream or TAO, or edge-inference optimization using TensorRT or ONNX
  • Experience with scalable training using Ray, distributed data loaders, or sharded checkpoints
  • Strong software engineering practices, including testing, linting, profiling, containerization, and reproducibility
  • Public technical work such as GitHub projects, first-author publications, or substantial open-source contributions

Related Jobs