ML Engineer (Fraud Detection & Data Quality)

TL;DR

ML Engineer (Fraud Detection & Data Quality) (Computer Vision/Adversarial ML): Building production systems that detect fraudulent, duplicated, manipulated, and copyright-risk media across millions of daily uploads with an accent on computer vision, vector search, and data-quality infrastructure. Focus on designing adversarial detection pipelines, clustering fraud networks and devices, and deploying human-in-the-loop verification at scale.

Location: Remote within the United States

Salary: $150,000–$250,000 base salary, plus $150,000–$350,000 in equity.

Company

A startup building an opt-in human data network that collects rights-aware photos, videos, and documents for AI labs and enterprises.

What you will do

• Build AI-generated image and video detection systems for authenticating user-submitted media.
• Develop reverse image search, plagiarism rejection, copyright-risk detection, and duplicate fingerprinting systems.
• Design EXIF and metadata tampering detection pipelines.
• Build fraud-network and device-clustering systems using vector and perceptual search techniques.
• Develop human-in-the-loop verification workflows and production ML infrastructure.
• Ship scalable backend and detection services processing millions of uploads per day.

Requirements

• 3+ years of experience in computer vision or machine learning.
• Production ML deployment experience with PyTorch or TensorFlow.
• Strong SQL and PostgreSQL skills.
• Experience with vector search technologies such as FAISS, pgvector, or Pinecone.
• Image-processing experience with OpenCV and PIL.
• Ability to ship backend systems using TypeScript, Deno, or similar technologies.

Nice to have

• Deepfake detection experience.
• Experience building reverse image search systems.
• Copyright detection pipelines.
• Trust and Safety infrastructure experience.

Culture & Benefits

• Work on applied AI data infrastructure used by a large network of global contributors.
• Own core systems that influence data used to train next-generation AI models.
• Operate at high volume, with 1.5–3 million uploads processed daily.
• Move quickly and ship systems in an adversarial production environment.
• Receive equity in addition to the base salary.