Research at the intersection of AI, systems, and applied learning.

Designing intelligent architectures that govern critical system resources, whether those resources are computational capacity, data, or learning experience, to improve the reliability, efficiency, and trustworthiness of AI systems.

ADSL

The Adversarial Detection and Data Sanitation Layer investigates how to detect, assess, and contain poisoned experience before it influences online reinforcement learning.

  • Detection and look-ahead validation.
  • Replay intervention and harmful-update containment.
  • Interactive research hypotheses, results, and run playback.
Explore ADSL

Hardware-Aware Dynamic Resource Allocation

Scheduling research that considers real compute constraints: capacity, contention, locality, changing demand, and workload-specific placement needs.

  • Hardware-aware scoring and placement.
  • Policy separated from resource discovery.
  • Extensible strategies for changing workloads.
Repo