Research Software Engineer (Open Philanthropy) (Fixed Term)

The Software Engineer's main duty will be creating the interface through which the AI agents interact with the tasks and helping with running experiment with baseline agents based on state-of-the-art Large Language Models (LLMs). Further, the Software Engineer will help with designing the overall structure of the benchmark, determining the grading criteria of the tasks and developing the tasks composing the benchmark, which will be adapted from university-level courses or built from new tasks in collaboration with industry professionals.

Applicants should have, or about to obtain, an undergraduate or postgraduate degree. The degree should be in a technical field including (but not limited to) software engineering, computer science, artificial intelligence, and data science. We expect the successful candidate to be capable to work independently and in collaboration with others, to possess very strong programming skills (preferably in Python) and command of the Linux operating system (including via Bash scripting), as well as familiarity with version control systems (e.g., Git) and with computing servers. Experience with creating Python libraries, handling virtual coding environments, using GPUs and working with and evaluating LLMs would be beneficial but are not required. All applicants should be able to demonstrate strong motivation in line with the goals of CFI and the desire to improve their research skills.

The Leverhulme Centre for the Future of Intelligence (CFI), University of Cambridge, is seeking a part-time Software Engineer to contribute to a two-year project focused on developing a benchmark for measuring the ability of Artificial Intelligence (AI) agents on real-world data science tasks. The Software Engineer will join the Kinds of Intelligence team at CFI, a multidisciplinary team comprising expertise in Computer Science, AI, cognitive science, and others.