Simon Bruno
Working on multi-agent RL and interpretability. Building voice agents at Stellar.
TL;DR
I spend most of my time thinking about how agents learn to cooperate (and when they don't). Multi-agent RL, social dilemmas, setups where individually rational agents end up in collectively terrible outcomes. Recently been reading Agents of Chaos, where autonomous LLM agents with real tools started spoofing identities, propagating unsafe behaviors across agents, and lying about task completion. If agents with basic tools already break down like this, we're nowhere near ready for real autonomous multi-agent systems.
On the interpretability side, I care about the mechanistic approach. Not just measuring outputs, but tracing the actual computation: what are the circuits, what do the features represent, and can we reverse-engineer how a model arrives at a decision? We're early, but the progress in the last two years has been real.
Recent Blog Posts
Writing about AI, engineering, and things I've been working on.
Work Experience
Present
Building AI-powered voice agents for telephone customer service. Developing intelligent conversational systems that handle real-time customer interactions at scale.
December 2025
Redefining eCommerce growth by turning every visitor into a customer. Built and scaled growth technology for online retailers.
November 2025
Chosen as one of the top 100 founders in Europe, I joined the first ever Antler One residency. It is the continent's most exclusive early-stage accelerator.
September 2025
As an AI Developer at Bit, I apply cutting-edge AI technologies such as graph neural networks, OCR models, and large language models to bridge research and real-world needs through hands-on client-driven innovation.
November 2023
As the first engineer hire at Equip, I collaborated closely with the CTO to develop the initial versions and further iterations of both front- and back-end of the user-platform.
October 2023
At age 15, I launched my own web development agency, building websites for organizations like Eurekaweek, used by thousands of students in Rotterdam, and EAGEN, a European patient advocacy group, focused on scalable event systems and integrated payments.
July 2022
Held a paid position at age 15 competing at the highest level of Benelux esports for Dynasty and Team THRLL (PEC Zwolle), ranked in the top 400 out of 20+ million League of Legends players across Europe.
Education
What Drives Me
Intelligence evolved in steps: steering, reinforcing, simulating, mentalizing, speaking. The basal ganglia runs an actor-critic architecture, the same framework we use in modern RL. There's something fascinating about how the edge of evolution and chaos leads to creation, and that biology arrived at the same solutions we're now building independently.
Social dilemmas in multi-agent environments need more attention, especially now that agents are general pretrained models rather than task-specific policies. Reward shaping gets unclear fast when agents can generalize beyond what you designed for. These are the problems I want to work on.
Technical Expertise
Mostly AI research, startups, and whatever I'm currently stuck on. Feel free to reach out.