How to apply for Research Engineer

About Iliad

Iliad is a nonprofit umbrella organization dedicated to advancing applied mathematics research in AI alignment, running conferences, research incubation, and fellowship programs. Unlike typical AI labs, Iliad focuses on supporting external and internal researchers rather than building products, making it a hub for foundational alignment work. Working here means directly enabling high-impact research that shapes the future of safe AI.

About the role

As a Research Engineer at Iliad, you will bridge the gap between exploratory research code and reliable, scalable production experiments. You'll work closely with alignment researchers to implement, debug, and optimize ML experiments, ensuring that the code faithfully executes the intended computations. This role is critical for accelerating alignment research by making experiments reproducible and efficient.

A typical day

You might start by syncing with a researcher to understand a new experiment they want to scale, then spend the morning refactoring their exploratory code into a robust training script. After lunch, you'd debug a failure in a distributed run, using LLMs to suggest fixes but carefully validating each change. Between projects, you'd automate a repetitive data preprocessing step or improve logging for better reproducibility.

Who Iliad is looking for

• Has strong ML engineering skills, with experience taking research prototypes and turning them into robust, efficient pipelines.
• Is a meticulous debugger who can trace failures in unfamiliar code and verify that implementations match researchers' mathematical intentions.
• Is proficient in using LLMs for coding and debugging, but maintains a healthy skepticism and validates outputs against ground truth.
• Thrives in a collaborative research environment, can explain engineering decisions clearly to non-engineers, and proactively improves tools and automation.

Tips for this application

• Highlight specific instances where you refactored or scaled research code (e.g., from a Jupyter notebook to a distributed training job) and the impact it had on research velocity.
• Demonstrate your ability to understand and debug unfamiliar research code by describing a time you had to reverse-engineer a complex model or algorithm.
• Show familiarity with AI alignment concepts—mention any relevant papers, projects, or contributions, even if not directly in alignment.
• Emphasize your experience with LLM-assisted coding: give examples of how you use tools like Copilot or ChatGPT effectively while catching their mistakes.
• Tailor your cover letter to Iliad's mission: explain why supporting alignment research appeals to you more than building products, and how you can enable researchers.

What to cover in your cover letter

Your cover letter should emphasize: (1) your track record of turning research code into production-quality experiments, with concrete metrics; (2) your debugging and verification skills, especially in unfamiliar codebases; (3) your ability to collaborate with researchers and communicate technical decisions; and (4) your motivation for supporting AI alignment research at a meta-organization like Iliad.

Research before applying

• Read Iliad's website and recent blog posts to understand their research incubation and fellowship programs, and the specific alignment projects they support.
• Familiarize yourself with key AI alignment research agendas (e.g., from MIRI, ARC, or academic labs) to speak the language of the researchers you'll support.
• Look into the backgrounds of Iliad's team and affiliated researchers to understand their technical focus areas and potential projects.
• Explore common tools and frameworks used in alignment research (e.g., PyTorch, JAX, Weights & Biases) and be ready to discuss your experience with them.

Likely interview topics

Based on the job description, expect questions about:

• How you would approach optimizing a slow training loop in a research codebase you've never seen before.
• Your process for verifying that an implementation matches a researcher's intended computation, especially when the math is complex.
• Strategies for using LLMs to accelerate coding while ensuring correctness and avoiding subtle bugs.
• A time you had to explain a technical trade-off to a non-engineer researcher and how you handled it.
• Your experience with maintaining research tools and automating repetitive tasks, and how you prioritize such improvements.

Common mistakes to avoid

• Focusing only on your ML engineering skills without connecting them to enabling alignment research—Iliad wants someone who cares about the mission, not just the tech.
• Overemphasizing your ability to write code from scratch rather than your skill in understanding, debugging, and scaling existing research code.
• Neglecting to mention how you verify correctness and catch errors when using LLMs, which could raise concerns about your rigor.