Applied computer vision for the environments that break most models.

I'm Faisel Mushtaq, a computer vision and an AI researcher. My work spans maritime intelligence, edge medical AI, and general perception problems — anywhere a model has to hold up under low compute, poor connectivity, or degraded input. I take on new domains readily; the constraints are usually more similar than the subject matter.

About

Applied computer vision, wherever it's needed

Based at the Center for Applied AI Systems. Previously trained in electrical engineering and machine learning.

Most of my work sits at the intersection of computer vision and constrained deployment: models that have to run on a handheld medical device with no signal, on a shoreline camera in fog, or on a satellite feed full of cloud cover. The domains I've worked in most — maritime intelligence and edge medical AI — happen to be where those constraints show up hardest, but the underlying skill set is general-purpose computer vision and applied ML, and I take on new problem areas as they come up.

I build and test against real deployment conditions rather than curated benchmarks, since that's where models actually have to work.

Projects

Selected work

Filter by area, or browse everything. Each project links to a fuller write-up.

Silicosis

Early detection of silicosis from chest imaging on portable X-ray hardware, aimed at mining and quarry communities where specialist review is scarce.

Medical imagingEarly detectionEdge inference

Urinalysis

Smartphone-camera-based urinalysis that reads standard test strips and reports results without a lab reader, calibrated for variable ambient lighting.

Point-of-careComputer vision

Heart Auscultation

Classifies heart sounds from a low-cost digital stethoscope, distinguishing common murmurs from normal variation directly on the recording device.

Audio MLScreening

Future Trajectory Prediction

Estimates where a vessel is likely headed, when it will arrive, and which ports it may visit after it disappears from AIS.

Trajectory forecastingETAPort prediction

All-Weather Satellite Image Enhancement & Ship Detection

Restores satellite imagery degraded by cloud, haze, and rain, and detects and classifies ships directly within the same pipeline, at scale.

Image restorationShip detectionRemote sensing
Cloudy / hazy Restored

Illustrative comparison — drag to compare.

Real-Time, Camera-Agnostic Distance Estimation & Ship Classification

Estimates vessel distance from any camera in real time — no calibration, no stereo rig — while classifying the ship in the same pass.

Monocular depthCamera-agnostic

Route Extraction & Real-Time Anomaly Detection

Learns normal vessel movement patterns from historical AIS data and flags real-time deviations, loitering, and emerging route changes.

Anomaly detectionAIS data
Publications

Selected papers

Interested in collaborating, or have a problem in a domain I haven't touched yet?