Production systems, hackathon winners, and deep learning research. All of it shipped.
Experience
Kissht
AI Engineer
01
Building AI systems on Kissht's consumer lending platform. Real shipped outcomes land here as the term goes on.
02
Working across the ML lifecycle: data pipelines, model iteration, and taking models to production.
Selected projects
HackCanada Sponsor Track Winner
Op. 01
PulmoScan
Flat CT scans, reconstructed as living 3D lungs.
An ML-driven lung diagnostic suite that turns 200+ flat CT slices into an interactive 3D model of the patient's actual lung. It goes after the 20 to 30 percent miss rate on actionable nodules and the 11-week referral gap that follows.
200+
2D CT slices fused into one 3D lung
20-30%
nodule miss rate targeted
11 wks
referral gap addressed via visual triage
CTRLHACKDEL 2.0 Overall Winner
Op. 02
NeuroSketch
A two-minute sketch that screens for Parkinson's.
A pre-Parkinson's screening platform that reads hand-drawn spirals and waves for early warning signs, squeezing a 15-minute clinical motor assessment into a 2-minute self-test.
<2 min
self-test vs 15-minute clinical assessment
400%
potential increase in triage throughput
CNN
MobileNetV2 with Grad-CAM explainability
Original composition
Op. 03
Satellite Change Detection
Two photos of Earth, and everything that changed between them.
A deep learning app that spots what changed between two satellite photos of the same place: new buildings, cleared forest, anything that moved.
0.84
F1-Score on change detection
1GB+
GeoTIFF files processed per pass
13
spectral bands normalized in the pipeline
Toolbox
Skills & technologies.
Languages
Machine Learning & Data
Web & Frameworks
Tools
Currently
These days I build AI for consumer lending at Kissht, and I'm open to future co-op terms. Building something ambitious? Let's talk.
Get in touch