Lead Research Scientist - Post-Training AI Agents

DeepMind Technologies Limited; Mountain View (CA); On-site; USD 262,000 - 364,000; Full time

Benefits offered by this job

Health insurance

401(k) with company match

PTO 20 days/yr

Sick time 40 hrs/yr

Maternity leave 28-30 weeks

Baby bonding leave 18 weeks

13 paid holidays

Job summary

Google DeepMind in Mountain View is seeking a Research Engineer/Research Scientist focused on core modeling and post-training experiences, including deep learning, LLMs, and reinforcement learning. You will build Consumer Agents and scale foundational post-training methods to convert foundation models into autonomous, adaptive agents for users.

This role emphasizes agent post-training, personalized context, data creation, and evaluation frameworks; you will create training datasets and test

Qualifications

• Bachelor's degree or equivalent practical experience.
• 8 years of software development experience.
• 5 years in design/architecture and launching software products.
• Experience with large language models, reinforcement learning, and deep learning.

Responsibilities

• Specialize in core modeling and post-training experience for agent post-training.
• Design and execute ablation studies to validate research hypotheses.
• Communicate findings and progress to the broader team.
• Develop research infrastructure and utilities for data analysis and model evaluations.
• Collaborate with other researchers and engineers to maintain feedback loops.

Skills

Large language models

Reinforcement learning

Deep learning

Software architecture

Testing/launching software

Education

Bachelor's degree or equivalent practical experience

Master's/PhD in Engineering or CS

Job description

Google DeepMind in Mountain View is seeking a Research Engineer/Research Scientist focused on core modeling and post-training experiences, including deep learning, LLMs, and reinforcement learning. You will build Consumer Agents and scale foundational post-training methods to convert foundation models into autonomous, adaptive agents for users.

This role emphasizes agent post-training, personalized context, data creation, and evaluation frameworks; you will create training datasets and test