It's all about the 1s and 0s
I enjoy building things that people find useful.
I am a Principal Engineer at Hive Learning, where I've worked since 2013. I joined as the first hire, building the platform for a coaching and learning tool shaped in its early years by Sir Clive Woodward. From there we expanded to build Hive Perform, AI-enabled coaching for sales teams, and in 2026 launched Kento, which builds AI agents for companies.
Along the way I built much of the original Python backend, a recommendation engine that is still running seven years on, the first generative-AI products, and now agents that enterprise sales teams use every day.
For me the technology is a means to an end, and I'm happy to learn whatever a problem needs.
September 2026: Side project: open-sourced mealie-toolkit, a local-LLM service that repairs and structures imported recipes.
August 2026: Wrote Why does your agent sound so confident when it's wrong?, introducing the Kento Canon.
May 2026: Hive Learning launches Kento, building AI agents for companies.
April 2026: Side project: released a Zigbee electricity meter on nRF52840/Zephyr, the successor to metermon.
November 2025: Took Hive Perform Labs into production, including an MCP server over the product.
September 2024: Started Hive Perform Labs: LLM analysis of sales calls for rep coaching and deal tracking.
February 2023: Built Hive's first generative-AI assistant on GPT-3.5, which grew into the Buzz Curator, Facilitator and Insights products.
January 2019: Promoted to Principal Engineer at Hive Learning. Our personalised recommendation engine went into production.
January 2017: Promoted to Senior Software Engineer; started the v4 backend.
January 2013: Started at Hive Learning (a.k.a. Coachbook, Captured).
March 2011: Started at QinetiQ.
November 2010: PhD Viva passed with minor corrections!
Technology Consultant, and technical lead on a programme researching emerging technologies for military simulation and training.
In 2010 I completed a PhD in Artificial Intelligence and Robotics at Imperial College London, in the Personal Robotics lab supervised by Prof. Yiannis Demiris.
I predicted the goals and intentions of teams of agents purely from observation, using a generative approach based on the simulation theory of mind. I built a 3D real-time strategy simulator, and ran headless faster-than-real-time copies of it across machines to test hypotheses.
For more information see the thesis page.
Partial Observability During Predictions of the Opponent’s Movements in an RTS Game, S. Butler and Y. Demiris, in Proceedings of the 2010 IEEE Conference on Computational Intelligence and Games, pp 46-53, IEEE, August 2010.
Using a Cognitive Architecture for Opponent Target Prediction, S. Butler and Y. Demiris, in Proceedings of the Third International Symposium on AI & Games, pp. 55-61, SSAISB, April 2010.
Multi-Agent Behaviour Segmentation via Spectral Clustering, B. Takacs, S. Butler and Y. Demiris, in Proceedings of the AAAI-2007 Workshop on Plan, Activity and Intention Recognition (PAIR), pp. 74-81, AAAI Press, July 2007.
Predicting the movements of robot teams using generative models, S. Butler and Y. Demiris, in Distributed Autonomous Robotic Systems 8, pp. 533-542, Springer, May 2009.
Operationalising the Simulation Theory for Intent Prediction in a Multi-Agent Adversarial Environment, PhD Thesis, Imperial College London (University of London), November 2010. (More info)