Research Engineer, AI Safety & Alignment

Written by the T&S Network from Character.AI's posting.

Research engineering on AI safety and alignment for large language models, covering evaluation, adversarial testing, and mitigation of harmful model behavior. The work sits on the Safety team and turns research into safety features shipped in the company's consumer chat product.

• AI misuse and jailbreaks

What you'd do

• Build new evaluation methods and metrics to measure safety and alignment of large language models.
• Run adversarial testing to find vulnerabilities and failure modes in models.
• Reduce bias, toxicity, and other harmful model behaviors using RLHF and fine-tuning.
• Research alignment, value learning, and interpretability techniques for models.
• Work with engineering and product teams to make safety solutions practical and scalable.
• Write production-facing and training code to test hypotheses and integrate safety fixes into products.

What it asks for

• PhD or equivalent experience in computer science, machine learning, or a related field.
• Ability to write clean production-facing and training code.
• Experience working with GPUs for training, serving, and debugging.
• Familiarity with the systems that move and store data for ML work.
• Strong understanding of transformers and reinforcement learning, including their safety implications.
• Background running product experiments and A/B tests (preferred).
• Hands-on work training large models across multiple machines (preferred).

Terms in this role

RLHF.

Reinforcement learning from human feedback: training an AI model on people's ratings of its answers.

Our best summary of the posting, and it can be wrong. The employer's posting is what counts: read it in full on Character.AInorth_east before you apply. If we got something wrong, report this listing.