Research Engineer

Responsibilities

• Develop and train scalable interpretability assistants to predict and detect subtle behaviors from model activations
• Create diverse evaluations by identifying interesting and undesirable behaviors in open-source models
• Develop novel architectures and objectives for training interpretability assistants
• Scale training and inference pipelines to support models up to 1T parameters

Requirements

• Experience fine-tuning language models and designing new architectures
• Proven ability in creating evaluations and experimental design
• Strong programming ability with a focus on balancing prototyping speed and maintainability
• Experience producing reliable results with epistemic self-awareness and transparency
• Strong communication skills and a collaborative, low-ego mindset

Preferred Qualifications

• Deep curiosity regarding the inner workings of ML systems
• Ability to generate original, productive ideas to unblock research progress

Benefits

• Competitive salary range of $250,000 - $500,000 per year
• Comprehensive benefits package
• Opportunity to work at the frontier of AI research with high direct impact
• Visa sponsorship available for international candidates

About the Company

Transluce is a non-profit research lab building tools for scalable, end-to-end oversight of AI systems. We develop world-class, AI-backed analysis tools and set industry standards for evaluation. Our work is integrated with core agent benchmarks and directly informs global regulation, including serving as a main evaluation developer for the EU AI Office regarding harmful manipulation risks.