Member of Technical Staff - ML Security Engineer

Role Description

Elloe’s safety loop only works if we can simulate attacks before they reach real users. You’ll drive red teaming at the system level — from jailbreaking to fuzzing and help shape the defense logic in AutoHeal and ReplayHeatmap.

What You’ll Build

• Red Team Simulation Engine
  • Launch adversarial attacks (prompt injections, bypass chains, logic traps)
  • Generate incident traces used to train AutoHeal patch logic
  • Contribute to fuzzing harnesses, risk scoring, and breach labeling
• Security Patch Forecasting
  • Forecast deployment risks based on incident graph trends
  • Automate “patch windows” based on SHAP mismatch clusters
  • Help convert audit traces into product-stoppable violations
• System-Level Defense
  • Build explainability-linked guardrails across the stack
  • Collaborate with infra and explainability leads on enforcement crossover

Qualifications

• Deep experience in adversarial ML, red teaming, or fuzz testing
• Understands how explainability can be a security surface
• Bonus: experience with threat modeling, diff-testing, or jailbreak detection

Why It Matters

Red teaming isn’t an afterthought. It’s how Elloe gets trusted to run in places that can’t afford to fail.

Why Now

Major regulators are asking for test results before they approve AI deployment. Institutions want defensible logs, not theoretical attacks. This is the moment when red teaming goes from research to required.

You’ll Leave This Role With

• A safety portfolio tied to real red team incidents, not just demos
• Impact on critical infrastructure used by hospitals, central banks, and governments
• A seat at the table defining what trustworthy GenAI actually means

Logistics & Application

• Start Date: Flexible (Q3 ideal)
• Location: Remote-first; timezone overlap with NY or DC preferred
• Comp: Competitive salary + equity

To Apply: Share one failure mode you’d simulate in a next-gen red team harness.