Longer write-ups of each project: the question, what I did, and what came out.
How training-set sampling affects QCNN intrusion detection
Under a fixed training budget, spreading attack records evenly across types raises AUC but lowers recall at a 1% false-positive rate. Under review at ICAIIC 2027.
Write-ups of the decoding work for quantum error correction will be added once the papers are public.
Context-aware anomaly detection in single-person households
Daily routines from six ambient sensors, and anomaly detection within each person's own routine. Context changes which days get flagged, but a simpler global baseline scored higher. Three papers and a registered program.