London Software Guild #4
About the London Software Guild
The London Software Guild gathers developers, architects, and teams across London who care deeply about the craft of building great software. No fluff, no vendor pitches — just people who take the work seriously, sharing what they've learned the hard way.
Agenda:
Abstract: The AI agent is the medium through which software is now developed. This results in much higher volumes of code, creating a new bottleneck: reviewing the outputs. In this talk, we describe how we designed our SDLC for this new world and how, as a security company, we maintain its security. We'll discuss linting, graph traversal, slop radars and automated PR approvals.
🎤 Double Act: Speakers: Jack Miller — CTO & Co-Founder alongside Chris Wise — Founding Cybersecurity Engineer, @ HelmGuard AI
Jack is the CTO and Co-Founder of HelmGuard AI, where him and his team develop AI-native security solutions for enterprise risk teams. His research focuses on the safety and robustness of machine learning, including AI verification and adversarial robustness. Before HelmGuard, he was an ERA Fellow at Cambridge, specialising in AI governance, and conducted research at the Australian National University (ANU) and ETH Zürich.
Chris is a founding cybersecurity engineer, with over two decades of experience in cybersecurity, cloud architecture, and DevSecOps. He has previously held DevSecOps Architect roles at Cyberhaven and Aqua Security. At HelmGuard, Chris focuses on leveraging AI to enhance enterprise cyber-risk and third-party risk workflows. His primary interest lies in the practical intersection of technology, protection, and online harm.
Abstract: Move fast and secure things. Getting security out of the way of Engineers.
🎤 Speaker: David Read, CTO and Co-Founder of Ossprey Security
David Read is CTO and Co-Founder of Ossprey Security, a London-based startup building tooling that catches malicious open-source packages before they reach production. Previously a software engineering manager at G-Research, where he spent nearly six years protecting AI systems in finance. He started his career as a cyber-security specialist working on critical national infrastructure, and has spent more than fifteen years shipping production systems.
Abstract: Guillaume will create a live agent connected to over 1,000 tools via MCP, addressing key challenges such as context explosion, ambiguous routing, and security vulnerabilities. He will present solutions like dynamic tool discovery, which reduces the number of tools from 500 to 1-2, and code-mode execution. Each MCP tool is important as a trust boundary for security teams. This talk serves as a practitioner's guide to scaling MCP tools, applicable to any tool-hungry agent.
🎤 Speaker: Guillaume Lebedel, Co-founder and CTO @ StackOne
Guillaume Lebedel builds integration infrastructure for AI agents. Before this, he developed large-scale event and data architectures for machine learning-driven trading systems and B2B platforms.
He focuses on the practical reality of scaling agentic workflows. In his sessions, Guillaume demonstrates what happens when agents connect to thousands of tools, breaking down failures like context explosion and indirect prompt injection. He also writes regularly about these engineering challenges on the StackOne blog.