Open to AI-safety and ML roles Hyderabad, India --:-- IST

Mohammed Faisal Parvez

AI Safety & Machine Learning Engineer

Builds systems that decide when a signal is trustworthy enough to act on. Model auditing and interpretability, benchmarks that publish their failures, and bounded-autonomy controllers from GPU processes to accounts payable.

PROBE ACTIVE
0 1 1 false positive rate true positive rate
Decision Boundary Threshold (τ) 0.50
True Positive100%
False Positive100%
DiagnosticDRIFT
AUROC 1.00 — probe separated fine-tuned model from base perfectly. It was detecting fine-tuning drift, not a secret loyalty.

Lab · Aurora

Slow bands of light in the site's own colour language: blue where a result held, amber where it is unclaimed, red where it did not hold. The calmest of the eight and the easiest to read text over.

Source and licence
react-bits Aurora (GLSL, verbatim) on a raw-WebGL host. MIT + Commons Clause. Listed in LICENCES.md.
JavaScript on this page
11.6 KB gzipped for the effect, 79.5 KB in total. The home page carries 68.0 KB.
Reacts to
time only; the three colour stops are the site's blue, amber and red.
Off-screen, hidden tab, reduced motion
Paused, paused, one still frame.