Robot Learning Research Intern

Bosch USA

Pittsburgh, PAFresher3h agovia LinkedIn

Posted 18 Nov 2026, 01:34 ET

This job at a glance

Who can apply
New grads and up to 1 year
Must know
Machine learningPythonPyTorchNode.jsLinuxComputer vision
Nice to have
LLMs
Work
On-site in Pittsburgh · internship
Pay
Not stated by the employer.
Skill demand
Python55%Machine learning44%PyTorch28%Node.js3%of 99 similar entry-level jobs posted in the US this week
Listed
Nov 18, 2026 · still open

Sign in free to see which of these skills your resume already shows.

Company: Bosch USA

About the Role

The Robot Learning Lab at the Bosch Center for Artificial Intelligence (BCAI) in Pittsburgh seeks research interns to develop robots that acquire transferable skills, adapt to unfamiliar situations, and improve through real-world experience. The work is at the intersection of robot learning, multimodal machine learning, embodied AI, and generative AI, aiming to solve challenging robotics and automation problems with industrial impact. The goal is to develop reliable robotic systems that augment human capabilities and collaborate with people and other agents.

Research Opportunities Span

  • Robot foundation models: vision–language–action models, test-time adaptation and steering, and learning from deployment experience and failures.
  • Agentic robot systems: coding agents and multi-agent frameworks for skill discovery, composition, long-horizon planning, and robot control.
  • Dexterous and contact-rich manipulation: integrating vision, touch, and proprioception for precise interaction with rigid and deformable objects.
  • Transferable robot skills: cross-embodiment representation learning, learning from human and robot demonstrations, and retrieval-based policy generalization.
  • Simulation and world modeling: generative digital twins and Real2Sim2Real pipelines for policy learning, evaluation, and transfer to physical robots.

Key Responsibilities

  • Formulate compelling research questions, critically assess prior work, develop and implement new methods, and design experiments that produce clear, defensible scientific insights.
  • Define a focused research problem with mentors and collaborators.
  • Critically review relevant literature and articulate a testable research hypothesis.
  • Develop an R&D plan specifying proposed methods, data sources, experimental settings, evaluation metrics, milestones, and target publication venues.
  • Design and implement the proposed approach, maintaining reproducible code, documented experimental configurations, and clear records of results.
  • Evaluate methods in simulation and on physical robots through quantitative comparisons, targeted ablations, qualitative analysis, and theoretical analysis.
  • Share progress, relevant literature, experimental findings, and open challenges in weekly discussions.
  • Prepare a research manuscript with collaborators for submission to leading venues.
  • Contribute to invention disclosures and patent applications.

Requirements & Tech Stack

Successful Candidates Will Bring Strong Foundations In Machine Learning And Robotics, With Depth In One Or More Of The Following Areas

  • Generalization and transfer: learning multimodal representations and robot policies that transfer across tasks, objects, environments, and embodiments.
  • Policy learning and robot foundation models: imitation learning, reinforcement learning, generative policies, and the training, fine-tuning, or post-training of vision–language–action and other embodied foundation models.
  • Learning at deployment: test-time adaptation, policy steering, failure detection and recovery, and agentic frameworks for reasoning, skill composition, and continued improvement.
  • Data and simulation: combining human demonstrations, robot trajectories, and synthetic data; developing simulation or world-model-based pipelines for training, evaluation, and transfer to physical robots.
  • Effective implementation: translating research ideas into reliable software, implementing and debugging neural policies, and using efficient training, inference, and distributed GPU workflows.
  • Rigorous evaluation: selecting informative baselines and ablations, diagnosing failures and distribution shifts, examining dataset biases, and characterizing robustness, data efficiency, and generalization.

Qualifications

  • Currently pursuing a Ph.D., or a recent postdoctoral researcher, in Robotics, Machine Learning, Computer Science, Computer Engineering, Electrical Engineering, or a closely related field.
  • Minimum GPA of 3.0 on a 4.0 scale, or equivalent.
  • Strong foundations in machine learning and robot learning, including representation learning, policy learning, optimization, and experimental design.
  • Hands-on experience training or adapting modern robot policies and foundation models, such as vision-language-action (VLA) transformers or world action models (WAMs).
  • Relevant training experience includes multi-task pretraining or behavior cloning/supervised fine-tuning with action-token cross-entropy, diffusion denoising, conditional flow-matching, or joint video-action prediction objectives; or offline or online reinforcement learning post-training using actor-critic or policy-gradient methods.
  • Strong Python programming skills.
  • Experience implementing, debugging, and evaluating neural models using PyTorch or a comparable framework.
  • Hands-on experience using coding agents to develop, test, and maintain robotics and machine learning software.
  • Practical experience with physical robot platforms, including robot data collection, policy deployment, and closed-loop experimentation.
  • Ability to design rigorous evaluations using appropriate baselines, ablations, and diagnostic analyses.
  • Demonstrated research ownership, supported by publications, preprints, or substantial research projects.
  • Clear technical communication and effective collaboration.

Desired Qualifications

  • Experience training or adapting robot foundation models using heterogeneous human demonstrations, robot trajectories, and synthetic data.
  • Experience integrating large language models, vision–language models, or other multimodal foundation models into robot perception, planning, control, or evaluation workflows.
  • Experience with test-time adaptation, policy steering, embodied memory, failure detection and recovery, or agentic systems.
  • Experience in dexterous or contact-rich manipulation.
  • Experience with simulation, generative world models, digital twins, or Real2Sim2Real pipelines.
  • Experience with multi-GPU or multi-node training, efficient fine-tuning and inference, and reproducible workflows.
  • Proficiency developing robotics and machine learning software on Linux, using version control and maintainable software practices.
  • Theoretical depth in areas such as transfer learning, reinforcement learning, optimization, uncertainty estimation, or safe learning and control.
  • A strong publication record at leading robotics, machine learning, or computer vision venues.

Internships in the US · last 7 days

60 openings · 51 need no experience

Most hiring: Futurepillar, Amgen, American Water

See all 60 →

FirstHire never charges to apply, and real employers don't ask for money for tests, interviews or offer letters. How we list jobs · Report this job