How to apply for ML Research Resident (Q1 2025)

About Elicit

Elicit is a public benefit company building an AI research assistant that helps researchers make better decisions by prioritizing systematicity and transparency over outcome supervision. Unlike typical AI startups, Elicit focuses on legible reasoning and scalable ML systems, making it a unique place for those passionate about AI safety and interpretability. Working here means contributing to a product that could fundamentally improve how knowledge is advanced.

About the role

This 3-month ML Research Residency is a contract position focused on developing computational procedures that reliably improve a knowledge state over thousands of iterations. You'll design and test improvement operators that maintain stability over 1000 iterations while making genuine progress, starting with simple tasks like shallow refactoring of scientific papers and scaling to complex ones. The role is impactful because it directly addresses core challenges in AI transparency and scalable reasoning, with potential for real-world application in Elicit's product.

A typical day

A typical day might involve brainstorming and coding new improvement operators, running experiments to test their stability over thousands of iterations, and analyzing results to refine your approach. You'll collaborate with other researchers to ensure your work aligns with Elicit's transparency goals, and you'll document your progress for legible reasoning. Expect a mix of independent deep work and remote team discussions.

Who Elicit is looking for

• Has hands-on experience designing and testing improvement operators for knowledge states, ideally in a research or applied ML setting.
• Is comfortable working on both simple (e.g., scientific paper refactoring) and complex tasks, with a track record of scaling solutions.
• Demonstrates a strong interest in AI transparency and scalable reasoning systems, possibly through prior projects or publications.
• Can work independently in a remote, contract environment and is motivated by iterative, long-horizon research problems.

Tips for this application

• Highlight any experience you have with iterative improvement systems or operators that maintain stability over many iterations—this is the core of the role.
• Showcase projects where you worked on AI transparency or legible reasoning, even if they were academic or side projects.
• Emphasize your ability to start with simple cases (like text refactoring) and scale up to complex tasks; provide concrete examples.
• Tailor your resume to include keywords from the job description such as 'improvement operators', 'knowledge state', and 'scalable reasoning'.
• In your application, express genuine interest in Elicit's public benefit mission and how this residency aligns with your long-term goals in AI research.

What to cover in your cover letter

In your cover letter, focus on: (1) your experience with designing and testing improvement operators for knowledge states, (2) your understanding of AI transparency and scalable reasoning, (3) your ability to work on both simple and complex tasks with examples, and (4) why you are excited about Elicit's unique approach to systematic and transparent AI research.

Research before applying

• Read Elicit's blog posts and publications to understand their approach to systematic ML and transparency in AI research.
• Explore their product features, especially how they handle reasoning and knowledge improvement, to align your answers with their methodology.
• Look into the backgrounds of their team members, particularly those in research roles, to understand their research interests and potential interview focus areas.
• Familiarize yourself with key concepts in AI transparency and scalable reasoning, such as process supervision and long-horizon reasoning, to speak fluently during interviews.

Likely interview topics

Based on the job description, expect questions about:

• How would you design an improvement operator that maintains stability over 1000 iterations? Walk us through your approach.
• Describe a time you worked on a simple task (like refactoring) and scaled it to a complex one. What challenges did you face?
• What does 'legible reasoning over long horizons' mean to you, and how would you implement it in a system?
• How do you ensure transparency in AI systems, and what metrics would you use to measure it?
• Given Elicit's focus on process supervision, how would you balance making genuine progress with maintaining stability in a knowledge state?

Common mistakes to avoid

• Avoid generic applications that don't address the specific challenge of designing improvement operators for knowledge states.
• Don't focus solely on your general ML experience without connecting it to iterative improvement and stability over many iterations.
• Avoid showing lack of interest in AI transparency or scalable reasoning, as these are core to Elicit's mission and this role.