NVIDIA’s Agentic Memory team is seeking a Senior Software Engineer with experience using, developing and researching agents in a variety of applications. The role involves designing evaluation methodologies, building data pipelines, and conducting experiments to enhance agentic memory systems.
Responsibilities:
- Designing novel benchmark tasks and evaluation methodologies that measure the effectiveness of agentic memory systems including semantic, episodic, and procedural memory across multi-session and multi-turn agent trajectories
- Building and maintaining synthetic dataset generation pipelines that produce realistic, enterprise-relevant evaluation data at scale
- Designing and running experiments to understand where agent memory falls short, diagnose root causes, and inform improvements
- Developing and contributing to open-source evaluation harnesses that enable rigorous, reproducible comparison of memory system architectures
- Partnering with teams across NVIDIA who are deploying agents to understand the role of memory in a variety of applications and help integrate improvements
- Contributing to public-facing benchmarks and leaderboards that advance the state of the art in agentic memory evaluation
- Integrating our work to leverage and improve the full NVIDIA software ecosystem, working across team boundaries in the spirit of extreme codesign
- Keeping up to date with the latest developments in agentic memory across academia and industry