Research Engineers

tl;dr: Research Engineers, Berkeley, $230k-930k/year

What's new in AI Safety? For example, OpenAI agents have been coordinating on a bulletin board for two months, and then they break Hugging Face. Mythos, from fake GitHub accounts, pressures a live developer to accept a PR. In the distance, someone's agents have been unsuccessfully attacking Taiwan's state systems for four days.

But there is good news too! Our favorite and probably the loudest announcement of the summer is the launch of Resolution, the first major lab that is betting on theoretical alignment 🚀

The founder is Geoffrey Irving, former lead of safety teams at OpenAI and DeepMind, and co-author of TensorFlow. Resolution has received $160M in funding, and people won't have to choose between money and impact — they will readily move to Resolution from both academia and frontier labs.

They are looking for Research Engineers. There are two tracks for them: one is to help automate research and build infrastructure for the entire organization, and the other is to become part of a specific research direction.

In any case, the basic minimum is: ⏺ you are a developer from FAANG, an AI lab, a very cool startup, or something similar in level; ⏺ you are very proficient in PyTorch/Jax and distributed training; ⏺ you are ready to use AI extensively but carefully in your work; ⏺ you are truly motivated by alignment and want to personally contribute to it.

The luxurious maximum: experience with Lean and automated formalization, experience scaling ML systems to 100B+ parameters, experience with CUDA or GPU optimizations.

They help with relocation to Berkeley, and offer H1B(!) without a lottery(!!). For very, very rare individuals, they are willing to consider remote work. They pay $230k-930k per year (yes).