Why this role
Attackers now use AI to find and exploit vulnerabilities faster than any team can patch. Defense has to become AI-native too, and that starts with knowing, precisely, which vulnerabilities are exploitable where they run.
You'll own that judgement for Defendermate: what makes a vulnerability exploitable on a real system, how attacks chain through a cloud environment, and how we prove we're right. Then you'll encode it in agents that make the call at machine speed, for every team that uses us.
What you'll build
- The research behind every call we make: the conditions that decide whether a vulnerability is exploitable, grounded in source code, patch diffs and public exploits.
- The agents that apply it: AI agents that verify exposures in a customer's environment, from evidence, without attacking it.
- The proof: test sets and a public benchmark that show how often we're right, especially the costly mistake of calling something safe when it isn't.
- The team: you hire and lead our detection & mitigation and compliance engineers.
What you'll do
- Turn vulnerability research into precise, testable conditions and real-world test cases.
- Map attack techniques and how they chain across cloud identity, network and workloads.
- Review automated results for errors and feed the corrections back into the agents.
- Publish research, CVE teardowns and benchmark results; speak at conferences and on podcasts.
- Join technical evaluations with enterprise customers.
- Write agents yourself when that's the fastest way to encode your judgement.
Your first year
- First 90 days: the test sets and evaluation harness are running, our first benchmark is published, and your two engineering hires are made.
- First year: our error rate is measured and falling release over release, our attack-technique coverage spans the use cases we launch with, and our research is a regular, public voice.
You might be a fit if
- You've made real vulnerabilities work, or proven that they don't, and turned that into repeatable checks.
- You can reason about attack paths in the cloud, especially AWS: identity and role trust, privilege escalation, network exposure, Kubernetes.
- You treat "not exploitable" as a claim that needs proof.
- You write strong, tested Python and use AI tools heavily, owning every line you ship.
- You've led security engineers or a research team.
Bonus points: research at a scanner, exposure-management or attack-simulation company; public work such as CVE credits, talks (nullcon, BSides, Black Hat) or open-source tools; attack-graph modelling; MITRE ATT&CK; agent tooling.