Software Engineer, Applied AI Agents, Professional Intelligence, DeepMind

Google•Mountain View, CA

6d•$147,000 - $210,000

About The Position

At DeepMind, the mission is to build the world’s first general-purpose learning agent. This role works with AI agents developed by Machine Learning and Neuroscience research scientists, building systems to test agents across 2D and 3D games and physics simulators, visualize results, create agent leaderboards, and test algorithms on robots. The position calls for a strong software engineering foundation and an interest in varied, challenging problems within a mission-driven team. In the Applied AI Agents, Professional Intelligence role, you will build capabilities for LLM agents that search, read, reason over, and cite large private corporate information sources. You will make agent outputs verifiable through grounding and citation checks, access-aware retrieval, and automated review, and lead evaluation efforts that measure factual accuracy, citation fidelity, and multi-step reasoning. You will ship production agent systems with strong code quality, testing, CI/CD, and monitoring. DeepMind describes itself as a pioneering AI lab with interdisciplinary teams advancing AI to address global challenges and accelerate product innovation. It emphasizes widespread public benefit, scientific discovery, and safety and ethics. Pay is determined by factors including job-related skills, experience, and relevant education or training. The US compensation range is $147,000–$210,000, plus a 15% bonus target, equity, and benefits.

Requirements

Bachelor’s degree in Computer Science or equivalent practical experience.
Experience in Python or TypeScript with modern toolchains and frameworks.
Experience building and shipping applications powered by LLMs, including prompting, tool use or function calling, retrieval-augmented generation, and agent orchestration.
Experience designing evaluations for LLM or ML systems and using them to drive measurable quality improvements.
Experience owning production software end-to-end, from design and implementation through deployment, monitoring, and iteration based on user feedback.
Experience using coding agents to develop code and tests to specification.

Nice To Haves

Experience with agent frameworks and protocols or agentic coding environments.
Experience with search and retrieval systems and citation or grounding verification.
Experience building products for legal, finance, compliance, or other regulated, high-stakes domains, or working closely with domain experts.
Exposure to fine-tuning or reinforcement learning for LLMs, and experience on Google Cloud Platform.
Experience with agent safety, red teaming, and adversarial testing of autonomous AI systems.
Experience or research background in recursive self-improvement (RSI) architectures, automated self-correction, or self-evolving agent frameworks.

Responsibilities

Act like an owner by asking questions, proposing solutions, building consensus, and driving execution from concept to delivery.
Design and implement skills, plugins, and typed tools for LLM agents that search, read, reason over, and cite large private corporate information sources.
Build grounding and citation checks, access-aware retrieval, and automated review of agent-authored knowledge so results can be audited to their sources.
Build evaluation suites and datasets with subject matter experts; measure factual accuracy, citation fidelity, and multi-step reasoning; and use results to drive improvements.
Ship production agent systems with strong code quality, testing, CI/CD, and monitoring.