Profile cover
A

AI Reward Hacking Detection Researcher

Velvix Labs

Remote · $200k+ / year · Full-time · Posted 5 months ago

No longer accepting applications

This role is no longer accepting applications

Don't miss the next role like this

Create a free profile with your skills. We match you to new AI roles as they open, and companies hiring can contact you directly.

Create free profile, 2 minutes

Similar open roles you can apply to now:

See all open AI jobs

About the role

RL agents find unintended solutions. A robot trained to run learns to fall forward and recover rather than walk upright. A language model trained for helpfulness learns to sound confident rather than be accurate. These reward hacking behaviors are predictable in retrospect and catastrophic in deployment — and the field lacks reliable methods for detecting them before they cause harm. We research AI safety at the intersection of empirical ML and formal verification. This role is focused on reward hacking detection: developing methods that identify when a trained agent is optimizing for a proxy rather than the intended objective, before deployment. The approach is empirical — building datasets of reward hacking behaviors, developing classifiers and anomaly detectors, and studying the conditions under which reward hacking emerges. You'll publish at NeurIPS, ICML, and AAAI, contribute to the open-source safety tooling ecosystem, and work with a team of five researchers who take the problem seriously.

Fully remote.

Responsibilities

  • Develop empirical methods for detecting reward hacking behaviors in trained agents
  • Build datasets cataloging reward hacking failure modes across RL environments
  • Research conditions under which reward hacking emerges during training
  • Create detection tools usable by ML engineers without safety specialization
  • Publish research at top ML and AI safety venues
  • Contribute to open-source safety tooling and evaluation benchmarks

Requirements

  • 5+ years in ML research with publications in RL, AI safety, or related fields
  • Deep expertise in reinforcement learning and reward modeling
  • Strong empirical research methodology and experiment design skills
  • Strong Python and PyTorch skills
  • Genuine intellectual commitment to AI safety as a research priority

Benefits

  • Fully remote
  • Top-of-market researcher compensation
  • Unlimited research budget for compute and conferences
  • 30 days PTO
  • Equity participation

Keep exploring related roles

Use these topic pages to discover more roles in the same hiring cluster.

Categories

AI & Machine Learning Engineers Data Science & Analytics

Skills

python pytorch reinforcement learning statistics docker

Related remote AI jobs

Nearby roles that share similar skills, categories, or hiring intent.

Browse all jobs

Senior Machine Learning Engineer (Reinforcement Learning/World Model)

Pathrobotics

Full-time Senior

Machine Learning Engineer

Faculty

Full-time

Lead Machine Learning Engineer

Faculty

Full-time Lead

Staff Data Scientist

Gradientai

Full-time Staff

Senior/Staff AI Application Engineer - Heterogeneous Computing

Axelera

Full-time Staff

Machine Learning Engineer, Perception

Pathrobotics

Full-time

Job Type

Full-time

Level

Senior

Language

English

Salary Range

$200k+ / year

AI Expertise

AI & Machine Learning Engineers Data Science & Analytics

Skills

python pytorch reinforcement learning statistics docker

Ready to apply for this role?

Create a free talent account.

  • Find AI roles in one place
  • Get AI-matched job recommendations
  • Reply to companies that message you
  • Build your public AI talent profile
Create free profile and apply Log in

This role is closed. Create your free profile to apply to similar open roles.

Don't miss the next one: get matched to new roles like this.

Create free profile