Red Team Engineer

present of AI; New York (NY); On-site; USD 140,000 - 200,000; Full time

Job summary

Reality Defender in New York seeks a security-focused researcher to probe our own deepfake detectors by simulating fraudsters and turning weaknesses into fixes for the AI team. The candidate should have 4+ years in offensive security or adversarial ML, experience cloning voices and performing face swaps, and strong Python with PyTorch.

Clear reports and independent work matter as much as tech skill. Join a fast-moving environment building defenses for real-time media analysis and detection

Qualifications

• 4+ years in offensive security or adversarial ML.
• Experience cloning voices and performing face swaps.
• Strong Python programming with PyTorch.

Responsibilities

• Attack our own deepfake detectors to replicate fraudster techniques.
• Turn findings into actionable fixes for the AI team.
• Produce clear written reports and work with minimal direction.

Skills

Offensive security

Adversarial ML

Python

PyTorch

Voice cloning

Face swap

Job description

The role is about attacking the company's own deepfake detectors the way real fraudsters would, then turning each weakness into a fix the AI team can act on. A strong candidate has at least four years in offensive security or adversarial machine learning, has actually made cloned voices and face swaps, and writes solid Python with PyTorch. Clear written reports and the ability to work with little direction matter just as much as the technical skill.

How it fits the AI era

AI makes convincing fakes cheap and fast, so the work is a constant race to reproduce new generation tools and see whether detection still holds. Automation can run test batches, expand attack catalogs and track regressions across releases. Deciding which threats are likely to be used next, judging what a finding means for customers, and setting safe rules for handling synthetic media remain human calls.

How it builds AI experience

Someone in this seat works directly with the newest voice, video and image generation tools and with detection models across several media types at a company built around catching deepfakes. To build this experience beforehand, try open source voice cloning and face swap tools, test them against public detectors, and write up what got through and why.

Worth showing

• Voice cloning and face swap projects
• Python and PyTorch attack tooling
• Adversarial ML evasion testing
• Automated test harness building

This is our own read, written from the employer's posting and the company's Spotlight. The employer's posting is the source of truth for the role.

About Reality Defender

A New York company whose API and SDKs, per the company, scan live calls, meetings, images and video in real time to flag AI-generated deepfakes before they reach a bank, call centre or government system.