At Bosch, we shape the future by inventing high-quality technologies and services that spark enthusiasm and enrich people’s lives. Our promise to our associates is rock-solid: We grow together, we enjoy our work, and we inspire each other. Welcome to Bosch.

The Robert Bosch GmbH is looking forward to your application!

• As a Senior Foundation Model Engineer, you will identify and evaluate open-weight VLA backbones for robotics (e.g., Pi0, Xiaomi robotics 0, smolVLA) and decide where post-training and fine-tuning make the decisive difference for real-world tasks.
• You will develop the training playbook, which includes post-training methods (e.g., preference/RL styles like RLHF/RLAIF/DPO) and task-specific supervised fine-tuning on open datasets, as well as Bosch motion data.
• Additionally, you will design and conduct rigorous benchmarks against vendor-fine-tuned releases, define strong baselines and ablations, and publish "Benchmark Cards" that show clear deltas.
• You will bridge the Sim2Real gap by aligning simulation datasets with physical tasks, applying domain randomization/retargeting, and validating the results on hardware (e.g., NVIDIA Isaac Sim, Groot).
• Moreover, you will be responsible for the evaluation harness and fixed data splits to ensure at least 98% reproducibility of experiments with precise run tracking.
• You will work end-to-end by implementing, launching, and analyzing large-scale experiments, debugging errors, and optimizing training/inference throughput, cost, and reliability.
• Furthermore, you will cooperate with Bosch data and platform teams to access datasets and computing resources, and you will package and document models for internal users and external demos.
• You will fully embrace agentic AI by using autonomous agents and copilots to explore architectures, structure pipelines, automatically generate tests/reports, triage runs, and accelerate iteration.
• Last but not least, you will engage with the community, publish results, write technical articles (e.g., Hugging Face blogs) with marketing support, and contribute to open source where appropriate.

• Education: Bachelor or Master degree in Computer Science, Machine Learning, Artificial Intelligence, Robotics, or similar
• Experience and Know-how: several years of experience in adapting large multimodal/vision-language-action models; post-training (RLHF/RLAIF/DPO) and SFT; rigorous benchmarking and ablation design; large-scale distributed training; Sim2Real strategies and on-robot evaluation; profound knowledge of Python (C++ is a plus); MLOps and experiment reproducibility; clear technical documentation and community engagement
• Personality and Working Style: you are an experimental person with a high degree of personal responsibility, and you always communicate your results precisely and honestly; you successfully maintain a balance between development speed and scientific accuracy and take full ownership of your results from start to finish
• Languages: very good written and spoken German and English

https://www.bosch-ai.com
www.bosch.com/research

Secondment to Corporate IP Ventures (C/IPV), acting for Robert Bosch Robotics GmbH (a 100% TOGE of Robert Bosch GmbH), limited to 12 months.

Please submit all relevant documents (CV, certificates, and links to GitHub or kaggle account).

We offer flexible working models: from various part-time options to mobile working and job sharing. Feel free to contact us.

Diversity and inclusion are not just trends for us but are firmly anchored in our corporate culture. Therefore, we welcome all applications, regardless of gender, age, disability, religion, ethnic origin or sexual identity.

Need support during your application?
Kathrin Stipak (Human Resources)
+49 711 811 38015

Need further information about the job?
Jan Schneider‑Kraus (Functional Department)
+49 172 974 9480
Christian Übber (Functional Department)
+49 152 0349 1987