Carnegie Mellon University

Description

The Computer Science Department of Carnegie Mellon University has an exciting opportunity for a technical postdoc for work on making large language models (and similar systems) safe and reliable, especially in contexts where they are used as agents and have access to tools. Some directions that are particularly of interest:

• Multiagent approaches, for example by having one agent monitor another, and agents working in concert
• Multiobjective optimization approaches to enable safety by design
• Evaluation, for example evaluating the performance of an agent over time, evaluating it in complex multi stakeholder environments, etc.

The position is for one year; we plan to be able to support a second year conditional on satisfactory performance. 

Specifically of interest are:

• Prior experience working with LLMs
• Prior experience conducting rigorous experiments
• An interest in decision making contexts and applications.

Also valued (but not required) are:

• Background or in related topics such as optimization, multiagent systems, game theory, human subjects experiments, ...
• Background or interest in interdisciplinary collaborations crossing into for example policy, operations research, economics, philosophy, ...

Qualifications

Candidates are expected to have or be about to receive a PhD in computer science or a related discipline, and prior experience with the type of work of the position.

Equal Employment Opportunity Statement

Carnegie Mellon University shall abide by the requirements of 41 CFR §§ 60-1.4(a), 60-300.5(a) and 60-741.5(a). These regulations prohibit discrimination against qualified individuals based on their status as protected veterans or individuals with disabilities, and prohibit discrimination against all individuals based on their race, color, religion, sex, or national origin. Moreover, these regulations require that covered prime contractors and subcontractors take affirmative action to employ and advance in employment individuals without regard to race, color, religion, sex, national origin, protected veteran status or disability.