About Grubel

Grubel is a research lab that builds AI for high-stakes knowledge work, starting with law. Legal matters require specialized expertise, and each one involves its own facts, practices, parties, and institutions. General-purpose AI applies the same system to every legal matter, varying only what goes into a limited context window.

To meet the depth and diversity of legal work, we let every matter build its own AI: engineer the right data for the matter, adapt the model and agent to it, and track what improves. We combine the three into a matter-specific improvement loop that runs until the system delivers what the matter demands.

Grubel was founded in 2026 by Moritz Hardt, director at the Max Planck Institute for Intelligent Systems and formerly a UC Berkeley professor and Google Brain researcher; and Reinhard Heckel, professor of machine learning at the Technical University of Munich, on leave to build Grubel. The founding team includes engineers with PhDs in computer science and experience in foundation-model pretraining at Mistral and legal AI research at Thomson Reuters. The lab has offices in Munich and Tübingen.

Point Nine led an oversubscribed pre-seed funding round with backing from leading AI researchers and founders, including Jeff Dean, Chris Ré, Ion Stoica, and Harvey co-founders Gabe Pereyra and Winston Weinberg.

About the role

We are looking for a Founding Research Engineer or Research Scientist to develop the methods and systems that make matter-specific AI possible. You will work on data curation, agent environment design, model adaptation, training, and evaluation. We do not draw a sharp boundary between research and engineering: everyone contributes substantial code, owns experiments end-to-end, builds systems, and helps turn successful methods into products. Research is empirical and engineering intensive.

You will work directly with Grubel’s founders, engineers, legal experts, and enterprise partners. As an early member of the team, you will also help shape Grubel’s technical direction and research culture.

What you’ll work on

• Developing, testing, and evaluating methods for adapting models and agents to individual legal matters

• Designing benchmarks and evaluation tasks for complex legal capabilities

• Developing data curation strategies for scaling from evaluation tasks to training tasks

• Building infrastructure for training, inference, experimentation, and evaluation

• Working with legal experts and enterprise partners to turn legal workflows into agentic workflows

• Turning successful research methods into reliable product capabilities

You may be a good fit if

• You hold a master’s degree or PhD in machine learning, computer science, mathematics, or a related field, or have equivalent research or industry experience

• You have either a strong machine-learning research record or significant experience building large-scale machine-learning systems

• You have trained and evaluated language models or built agentic systems

• You are proactive, self-directed, and comfortable owning technically ambiguous problems

• You are curious about how legal work actually gets done and enjoy working with domain experts

Particularly relevant experience

Any of the following would be valuable, but we do not expect candidates to have all of them:

• LLM post-training, fine-tuning, or reinforcement learning experience

• Data curation, synthetic data, or human-in-the-loop data generation pipelines

• Evaluation harnesses and benchmark development

• Distributed training, inference, or experiment infrastructure

• Publications, open-source contributions, or deployed research systems

• Legal AI or another expert-intensive application domain

Logistics

• Munich or Tübingen. We value working together in person, prefer candidates based near one of our offices, and expect at least one day per week in the office. We plan to establish a Zurich office in 2027.

• Full time employment

• Competitive salary plus a stock option grant in Grubel, Inc., our US parent company.