About Us
Engineering AI for Practical Development
Our team has built software with AI for decades, from biological research to conversational interfaces. We've experienced firsthand how challenging it is to apply AI effectively in real-world development. Maestro is our solution to this problem, combining frontier AI capabilities with practical development tools. We focus on empowering developers to build robust, production-ready software while maintaining full control over their development process.
Our Team
Sean Ward
Co-Founder and CEO
Sean brings over 20 years of experience in startups and machine learning. He founded Relatable (internet-scale music ID and recommendation), was a UCL Research Associate in Bioinformatics and protein structure prediction, and founded Synthace (synthetic biology/biological experiment automation and ML) and Scale DX (robot-driven ultra-high throughput saliva-based COVID diagnostics).
Martin Szummer
Co-Founder and CTO
Martin has over 20 years of experience in ML research and startups. He holds a PhD from MIT AI Lab and was a researcher at Microsoft Research, focusing on early language model research. Martin co-founded VocalIQ, developing self-learning conversational voice UI. He also served as a Staff Research Scientist at DeepMind, contributing to cutting-edge AI advancements.
Sam Shapley
Member of Technical Staff
Sam holds a First Class MSci in Theoretical Physics from Imperial College London, completing his thesis on simulation of high-energy particle physics with GANs. During this research, he began using early foundation models extensively to help ideate and automate development. Sam has previously worked in data, and researching LLMs and agentic systems for Weights and Biases.
Maxime Robeyns
Member of Technical Staff
Maxime is a final-year PhD student in probabilistic machine learning. His research focuses on reinforcement learning, large language models, density estimation, and Bayesian deep learning. Maxime has worked on Bayesian reward models for LLM alignment and post-training LLMs for code generation.