Research Science Intern (PhD)

<p><strong>Datadog AI Research — Scholars Program with Carnegie Mellon University</strong></p> <p>Datadog AI Research (DAIR) is partnering with Carnegie Mellon University to support a small number of PhD students working on open research problems grounded by ongoing efforts at Datadog/DAIR.&nbsp;</p> <p>You will frame a problem, run your own experiments, and write up what you find, with compute and data at a scale most academic labs cannot provide.&nbsp; You will collaborate with colleagues working on the same questions.</p> <p>&nbsp;</p> <p><strong>The Lab And The Research:</strong></p> <p>DAIR is an industrial research lab motivated by practical challenges in observability and software operation: detecting and diagnosing failures, understanding complex production environments, and helping engineers operate software more effectively. The lab focuses on creating specialized foundation models, post-training and evaluating AI agents, and building frontier-scale machine learning systems. By combining fundamental research with Datadog's large-scale, real-world data and infrastructure, the lab develops new AI capabilities and translates them into practical systems with meaningful impact.</p> <p>Internship projects are shaped with your DAIR mentor and your CMU faculty advisor. You do not need prior experience with observability, monitoring, or infrastructure.</p> <p>&nbsp;</p> <p><strong>What You'll Do:</strong></p> <ul> <li>Own a research project end to end: framing the question, running the experiments, writing it up</li> <li>Work directly with a DAIR mentor engaged in the same problem, and stay connected to your advisor and lab</li> <li>Publish, and use the work toward your dissertation</li> <li>See research reach production, when it works</li> </ul> <p>&nbsp;</p> <p><strong>Who You Are:</strong></p> <ul> <li>Currently enrolled in a PhD program at Carnegie Mellon in machine learning, computer science, statistics, or a related field</li> <li>Depth in at least one area relevant to the research above</li> <li>Comfort running real experiments — training models, working with GPUs, reading and reimplementing recent papers</li> <li>Evidence you can do research: conference or workshop papers, preprints, open-source contributions, or a lab project you can discuss in depth</li> <li>Research taste. You can explain why a problem matters and what would change if it were solved.</li> <li>Honest empiricism. You report the ablation that didn't work.</li> <li>Support from your faculty advisor.</li> </ul> <p>&nbsp;</p> <p><strong>Not Required:</strong></p> <ul>