Leading the AI revolution
in software development
From building one of the first agentic issue remediation systems to creating enterprise-grade AI coding harnesses, I work at the intersection of artificial intelligence and developer productivity.
AI is not replacing developers.
It is amplifying them.
The future of software development is not about AI writing code in isolation. It is about AI agents working alongside developers as intelligent collaborators, understanding context, enforcing quality standards, and accelerating the entire development lifecycle.
Throughout my career, I have focused on this vision: building AI systems that genuinely understand the developer workflow. Not superficial code completion, but deep integration into how software is conceived, built, tested, and maintained. At Sonar, I lead the AI Agents and Developer Tooling organization, where we are creating tools that make agentic coding practical for the world's largest engineering teams.
My conviction is that agentic coding will be the default mode of software development within a few years. The leaders who prepare their organizations now, who build the right infrastructure, evaluation pipelines, and governance frameworks, will have an enormous competitive advantage. The ones who wait will struggle to catch up.
Building the infrastructure for AI-powered development
Agentic coding goes far beyond autocomplete. It is about creating AI systems that can reason about codebases, identify issues, propose and validate fixes, and integrate seamlessly into existing developer workflows. Here is how I have contributed to making this real.
MCP Server Development
I built Sonar's first Model Context Protocol server, enabling AI agents to interact natively with SonarQube and SonarCloud. MCP is the emerging standard for how AI agents communicate with developer tools, and getting this right was critical. The server allows any MCP-compatible AI agent to query code quality data, retrieve analysis results, and take action on findings, all through a standardized protocol. This was not just a technical achievement but a strategic one: it positioned Sonar at the center of the agentic coding ecosystem.
Agent-First CLI
I reintroduced the Sonar CLI as an agent-first tool, fundamentally rethinking what a command-line interface means when AI agents are the primary users. Traditional CLIs are designed for human interaction. Agent-first means structured output, deterministic behavior, and seamless integration into automated workflows. The result is an enterprise-grade agentic coding harness that allows organizations to embed code quality directly into their AI development pipelines.
LLM Code Remediation
At Snyk, I built AI Codefix, one of the industry's first agentic issue remediation systems. This was a large language model trained to understand security vulnerabilities in code and generate precise, contextual fixes. Not generic suggestions, but production-ready patches that respected the codebase's existing patterns, dependencies, and security posture. This work required deep collaboration between product, AI research, and security engineering teams.
AI Evaluation Pipelines
Leading R&D on SonarEval, an AI evaluation pipeline that measures how effectively AI agents produce quality code. As AI-generated code becomes more prevalent, the industry needs rigorous ways to evaluate it. SonarEval addresses this by creating benchmarks and automated assessment workflows that measure not just whether code works, but whether it meets enterprise quality, security, and maintainability standards.
Guiding enterprises through the agentic coding transition
Adopting AI in software development is not simply a matter of buying tools. It requires rethinking workflows, establishing governance, training teams, and building the infrastructure to measure AI effectiveness. I work closely with enterprise leaders worldwide to design and implement best practices for agentic coding adoption.
This work spans the full spectrum: from C-suite strategy conversations about AI readiness to hands-on technical workshops with engineering teams. The enterprises I work with range from global financial institutions to major technology companies, each with unique constraints around security, compliance, and engineering culture.
A critical insight from this work is that successful AI adoption is not about the AI itself. It is about the surrounding infrastructure: the evaluation pipelines that measure quality, the governance frameworks that ensure compliance, and the developer experience that makes AI adoption feel natural rather than forced. When these elements are in place, adoption accelerates dramatically. At Sonar, we have seen 5x month-over-month growth in AI agent adoption by focusing on exactly this approach.
Principles for AI in developer tools
After years of building AI-powered developer tools across security, code quality, and developer experience, I have developed a set of principles that guide my approach.
Quality over speed. AI-generated code that introduces technical debt, security vulnerabilities, or maintainability issues is worse than no AI at all. Every AI system I build starts with measuring quality outcomes, not just throughput. This is why evaluation pipelines like SonarEval are foundational, not afterthoughts.
Agents, not assistants. The real value of AI in development comes from autonomous agents that can take meaningful action, not chatbots that answer questions. This means investing in agent infrastructure: MCP servers, structured CLI tools, and evaluation frameworks that allow agents to operate reliably at scale.
Developer trust is earned. Developers are rightfully skeptical of AI tools that promise too much. Building trust requires transparency about what AI can and cannot do, consistent reliability, and measurable improvements to their workflow. The PLG motion I drove at Snyk, reaching over seven million developers through education, taught me that developer adoption is built on genuine value, not marketing.
Enterprise-grade means governance. AI adoption in large organizations requires more than good models. It requires audit trails, compliance frameworks, configurable policies, and clear accountability. The tools I build are designed for regulated industries where moving fast does not mean moving recklessly.
Ready to lead your
AI transformation?
Whether you are exploring agentic coding adoption, building AI evaluation pipelines, or rethinking your developer tooling strategy, I would love to connect.