Sholto Douglas

Sholto Douglas works on scaling reinforcement learning at Anthropic, including infrastructure for experimentation and research. Previously, he was a research engineer at Google DeepMind; his earlier research focused on robotics at the University of Sydney’s Australian Centre for Robotics. In public conversations, he discusses how reinforcement learning, model capabilities, and reliability shape the development of AI agents.

Career

Cosigned byLoading...

CosignedSign in to view

Based in: San Francisco
Elsewhere

Anthropic

Feb 2025 — Present

Member of Technical Staff

Google DeepMind

Research Engineer
Sep 2022 — Jan 2025
Research Engineer
Sep 2022 — 2025

Australian Centre for Robotics

2019 — Dec 2019

Honour's Researcher

JD.COM

Apr 2018 — May 2018

Intern

Zeroth.ai

Feb 2018 — May 2018

Intern

Tsinghua University

Bachelor's studies

Udacity

The University of Sydney

Bachelor's studies

Products & media

dwarkesh.com

Is RL + LLMs enough for AGI? — Sholto Douglas & Trenton Bricken

Sholto Douglas and Trenton Bricken discuss reinforcement learning, mechanistic interpretability, and the practical limits of autonomous agents with Dwarkesh Patel in this May 2025 interview transcript.

Media appearance

dwarkesh.com

How LLMs actually think — a written interview

An edited 2024 conversation between Sholto Douglas, Trenton Bricken and Dwarkesh Patel about long-context models, reliable agents, inference systems and interpretability.

Writing