Senior security engineer, hands-on across cloud, application, and AI security. I secure AWS, wire security into CI/CD, threat model before code ships, and I go deep on authorization and multi-tenant SaaS risk. I built Eidolon, an open-source security agent, and I lead Canada's largest hacker community.
I'm a senior security engineer who works hands-on across cloud, application, and AI security. I secure AWS, IAM and least-privilege, secrets, and misconfiguration, I wire SAST, dependency, and container scanning into CI/CD without the noise that makes developers ignore it, and I threat model new products and APIs before they ship.
My specialty is authorization and multi-tenant SaaS risk, which for a platform handling sensitive financial data is most of the game, backed by 120+ validated vulnerabilities through HackerOne. On the AI side, I review LLM and agent workflows for data leakage, prompt injection, and excessive permissions, and I built Eidolon, an open-source orchestrator that runs security workflows with an agent under scope tokens and command gating, because an agent with more access than the user is the whole risk.
I have also led incident response end to end, from investigation and containment through root-cause and runbooks, and I build reusable tooling so a control gets written once and reused everywhere. I lead DEF CON Toronto (DC416) and speak at SecTor and DEF CON Vancouver.
A trust boundary drawn wrong on a whiteboard costs an afternoon. The same mistake in production costs a quarter. I model abuse cases with the engineers who wrote the doc.
A scanner at default settings buries a team in noise until they stop reading it, which is worse than nothing. I'd rather ship five findings a week that are all real than five hundred that aren't.
Findings get traced to root cause, then checked for the same pattern everywhere else. One IDOR is a bug. The same authorization mistake in nine places is a design problem.
A Burp Suite extension I built that automates OAuth2.0 and OIDC authorization-bypass detection across API surfaces. Auth is where identity systems break, so I made the bug class easier to find.
Open source (MIT), built and maintained in Python. A security orchestrator that automates testing by driving an AI agent, with per-engagement isolation, scope tokens, and three-tier command gating, so the agent earns trust one step at a time. github.com/amir-hosseinpour/eidolon →
SecTor 2025 and DEF CON Vancouver (API attack chains and OAuth2.0 exploitation), plus a firmware teardown of a robot vacuum under the vendor's bug bounty program. I share the work rather than hide it.
Happy to talk whenever works for you.