Staff Research Scientist - Post-Training for Agents

At Datadog in the Technology space, Observe, Secure, and Optimize Modern Cloud Infrastructure at Scale.

<p> </p>

<p>As a Staff Research Scientist within Datadog AI Research (DAIR), you will drive research in post-training and autonomous agents as a hands-on individual contributor. You will advance reinforcement learning, simulated environments, synthetic data generation, and evaluation methods for agents operating across complex real-world systems. You will set the technical direction for ambitious research programs, raise the technical bar for the researchers and research engineers around you, and collaborate with Datadog's product and engineering teams to translate research advances into products. </p>

<p> </p>

<p><strong>What You'll Do: </strong></p>

<ul>

<li>Drive research in agent post-training and reinforcement learning, shaping the technical direction of ambitious research programs grounded in observability and security</li>

<li>Own research problems end to end, from framing the question through experimentation, model development, and evaluation</li>

<li>Design and advance reinforcement learning approaches, post-training methods, and training loops for agents operating in complex environments</li>

<li>Build simulated environments, synthetic data generation approaches, and evaluation frameworks that enable scalable agent training and rigorous measurement</li>

<li>Raise the technical bar across the team by reviewing research directions, mentoring researchers and research engineers, and setting standards for experimental rigor</li>

<li>Collaborate with cross-functional teams across Research, Product, and Engineering to translate advances in autonomous agents into scalable Datadog capabilities</li>

<li>Contribute to research publications, present at top-tier conferences such as NeurIPS, ICLR, and ICML, and help open-source key model artifacts and benchmarks</li>

</ul>

<p>

Want to learn more about what it's like to work at Datadog? View our full profile.