Teams
The lab is small on purpose: each team owns its problems end to end, with a named lead, a scoped review cadence and the shared harness under every result.
Early-warning detection and sensor-level trust for vehicles, drones and industrial robots. Owns the pre-attack telemetry and threat-horizon problems.
Self-defending OS research around the Black Wall runtime: learned defence models, adaptive policy and cyber-physical containment.
Federated learning with formal privacy guarantees, differential-privacy accounting and deployment-grade quantization for edge hardware.
Adversarial robustness of code generation agents, semantic code search and the security audit tooling around AI-written software.
Lightweight SLAM, perception pipelines and control stacks running on the drone and microcontroller hardware the lab keeps in-house.
Placements are scoped to a single problem with a named supervisor and a deliverable. Remote, asynchronous, and open worldwide.
