Work with me

I help teams ship AI agents they can defend: find what really breaks, fix it in the design, and keep it fixed in CI. Remote, from Switzerland. Available for roles and contracts from November 2026.

AI agent security assessment

For a team about to put an agent in front of customers or data.

• Red-teaming of your agent over its real interfaces (REST, MCP, A2A, chat), including indirect injection through tools and knowledge bases
• Every finding with the exact prompt and reply, confirmed by deterministic checks rather than another model's opinion
• Fixes in order of impact: tool allow-lists, session-level business rules, identity and least-privilege access
• Evidence mapped to the EU AI Act, DORA, NIS2 and GDPR, ready for your risk and audit teams

Proof: the benchmark write-up, where this method caught what seven tools and an LLM judge missed.

Secure agentic platform architecture

For a team designing or reviewing an agent platform on Azure or Microsoft Foundry.

• Agent identity and authorization: on-behalf-of flows, non-human identities, scoped MCP tools
• AI gateway, content safety and evaluation design (Azure API Management, Content Safety)
• Architecture review with findings you can act on, not a slide deck

Proof: 14 years in defence, pharma and building-automation systems; Microsoft Multi-Agent AI Solutions Expert and Azure Solutions Architect Expert.

Red-teaming in CI

For a team that wants agent security checked on every change, not once a year.

• scan-action and sixi-scanner wired into your pipeline, findings in the GitHub Security tab
• Custom checks for your agent's own business rules: the ones generic tools can't know
• A handover so your team owns it afterwards

How it starts

A 30-minute call: what the agent does, what it can touch, and what would hurt. Then a fixed-scope proposal. Message me on LinkedIn.