How to apply for Machine Learning Engineer

About Elicit

Elicit is a public benefit company building an AI research assistant that searches over 125 million papers, extracts data, and surfaces insights for researchers, policy-makers, and industry leaders. Its approach centers on supervision of process, not outcomes: breaking complex questions into human-legible steps and supervising the reasoning process itself rather than relying on large black-box models. Hundreds of thousands of researchers have used Elicit for literature reviews and systematic reviews.

About the role

As a machine learning engineer, you will build products and workflows that help researchers and scientific teams make higher quality decisions with language models. The role involves working on a scalable ML system that prioritizes systematicity and transparency, and it is explicitly not for someone who only wants to develop models in isolation from users. You will help expand Elicit's impact beyond academic research while laying the groundwork for ML systems that reason at scale.

A typical day

A typical day likely involves building or improving workflows that use language models to search papers, extract data, or surface insights for users. You would work on a scalable ML system that breaks complex questions into human-legible steps and supervises the reasoning process. The exact daily routine is not described in the job details, so ask about team structure, standups, and how engineers collaborate with product and research during interviews.

Who Elicit is looking for

• Has built ML systems that prioritize systematic, transparent, human-legible reasoning steps over large black-box models.
• Is comfortable working on user-facing products and workflows, not just isolated model development.
• Has experience with language models applied to search, extraction, or insight generation over large document collections.
• Wants to work at a public benefit company where supervision of process, not outcomes, is a core design principle.

Tips for this application

• Read Elicit's roadmap and recent Twitter posts to understand how people use the product today, then reference specific use cases in your application.
• Show that you understand the difference between supervising process and supervising outcomes, and explain why that distinction matters for your work.
• Describe concrete ML systems you have built that involved human-legible intermediate steps, not just final model outputs.
• Demonstrate interest in the research and scientific decision-making domain, since Elicit serves researchers, policy-makers, and industry leaders.
• Avoid framing yourself as a model-only engineer. This role explicitly rejects that profile, so highlight product and workflow contributions.

What to cover in your cover letter

Explain why supervision of process, not outcomes, fits how you build ML systems. Describe a specific project where you broke a complex question into human-legible steps and supervised the reasoning process. Show familiarity with Elicit's product, such as literature reviews or systematic reviews over 125 million papers. State why you want to work on a scalable ML system that prioritizes systematicity and transparency.

Research before applying

• Use Elicit itself on a research question to see how the product handles literature reviews and systematic reviews.
• Read Elicit's roadmap to understand where the company is expanding beyond academic research.
• Read Elicit's Twitter posts to see how researchers and other users describe their workflows and needs.
• Look into the concept of supervision of process versus supervision of outcomes, and be ready to discuss tradeoffs.

Likely interview topics

Based on the job description, expect questions about:

• How you would break a complex research question into human-legible steps that can be supervised individually.
• Your experience building ML systems that search over or extract data from large document collections.
• How you balance transparency and systematicity against raw model performance in a production ML system.
• How you would design a workflow that helps researchers go from questions to evidence-backed decisions in minutes.
• Why supervision of process, not outcomes, is a safer path toward advanced AI, and how you would implement it.

Common mistakes to avoid

• Presenting yourself as only interested in developing models in isolation from users, which the job description explicitly rules out.
• Treating Elicit as a generic AI company instead of engaging with its specific focus on research, systematic reviews, and transparency.
• Claiming to value transparency without giving concrete examples of how you have built systematic, human-legible ML steps.