The opportunity
As a Research Engineer on our team, you will partner with Research Scientists to turn research ideas into working systems, building the data, tooling, and infrastructure that enable rapid iteration, trustworthy evaluation, and a smooth path from prototype to production.
What you'll do
World Models for Observability -: Training multimodal foundation models that learn the joint dynamics of distributed systems across metrics, traces, logs, topology, and events. These models power advanced forecasting, anomaly detection, root cause analysis, counterfactual simulation ("what if?"), and provide a learned planning backbone for our autonomous agents.
Trained Agents for Observability -: Post-training models to operate autonomously across Datadog's domain. SRE incident response is our first target, with a clear path to code repair, security response, and infrastructure optimization. We build the simulation environments, RL training loops, and evaluation infrastructure needed to train agents that match or surpass frontier models at a fraction of the cost.
Build and operate multimodal data pipelines, training and evaluation: infrastructure, benchmarks, and internal tooling
Implement models, run experiments at scale, and profile for reliability, performance, and cost
Build simulation environments and replay infrastructure for agent training and evaluation
Orchestrate distributed training and distributed RL with Ray, including: scheduling, scaling, and failure recovery
What they're looking for
- Establish rigorous automated benchmarks and regression tests for world model: predictions, agent performance, and simulation fidelity
- Collaborate with Research Scientists, Product, and Engineering to integrate: capabilities into Datadog's products and to harden prototypes into reliable services
- Contribute to research publications at top-tier conferences (e.g., NeurIPS,: ICLR, ICML), and produce high-quality code, documentation, and open-source artifacts
- You have depth in distributed computing, RL Infra, and ML systems for: training and inference at scale; experience with Ray, Slurm, or similar frameworks is a plus
