Agentic AI & Autonomous Software
Long-horizon agents that plan, call tools and verify their own work.
A student-driven research group working at the edge of emerging technology.
We pick problems a few years ahead of the curve, build a real prototype, and break it on purpose. Code first, slides last.
Every track starts with something running.
If it can't survive a red team, it isn't finished.
Undergraduates get commit access on day one.
# Code Research Lab from crl import Lab, Agent lab = Lab(tracks=6) for t in lab.tracks: swarm = [Agent(t) for _ in range(64)] result = t.run(swarm, adversarial=True) assert result.reproducible # >> 6 tracks online # >> status: recruiting
Each track is an active build lane with its own repo and bench.
Long-horizon agents that plan, call tools and verify their own work.
Real-time control loops and high-fidelity twins of physical hardware.
Dry-electrode EEG and EMG decoded in real time, cortex to actuator.
Lattice-based key exchange and migrating real systems before the deadline.
Headset-native interfaces, 3D reconstruction and Gaussian splatting.
SLAM, visual navigation and swarm consensus that survives dead nodes.
A scheduler for fleets of autonomous coding agents, with adversarial verification passes.
A hardware-in-the-loop rig that runs a digital twin against the real board and flags where they diverge.
Sub-10 ms dry-EEG acquisition and decoding, closed loop from cortex to actuator.
A drop-in hybrid ML-KEM transport layer, plus tooling to audit legacy cryptography.
Phone-captured scenes reconstructed as Gaussian splats and walked through in a headset, same session.
Multi-robot coordination that keeps consensus through packet loss and node death.
No GPA gate. Bring curiosity and we'll hand you a track, a repo and a bench.