AI Research Engineer

What you'll do

• Build and operate multimodal data pipelines, training infrastructure, and internal tooling for research models.
• Implement models and execute large-scale experiments while profiling for reliability, performance, and cost.
• Develop simulation environments and replay infrastructure to support agent training and evaluation.
• Orchestrate distributed training and reinforcement learning using frameworks like Ray.
• Establish automated benchmarks and regression tests for world model predictions and agent performance.
• Transition research prototypes into hardened, reliable services integrated into Datadog products.
• Contribute to research publications at top-tier conferences and produce high-quality open-source artifacts.

What we're looking for

• Proficiency in Python and familiarity with systems languages like Rust, C++, or Go.
• Experience with distributed computing, RL infrastructure, and ML systems for training and inference at scale.
• Practical experience operating ML training and inference systems using PyTorch or JAX.
• Expertise in large-scale model training and fine-tuning using frameworks like Megatron-LM, DeepSpeed, SkyRL, VeRL, or TorchTitan.
• Experience with techniques such as SFT, RLVR, RLHF, and efficient inference methods like quantization and speculative decoding.
• Familiarity with Ray, Slurm, or similar frameworks for orchestrating distributed training and RL.
• Experience building multimodal data pipelines, simulation environments, and evaluation infrastructure.
• Experience supporting or contributing to research publications at top-tier conferences.