Selected work

Systems in production, not slideware.

A slice of the AI work we've shipped — computer vision for sports analytics, security, and accessibility. Financial-software engagements are covered under NDA and shared on request.

Soccer stadium with players on the field
Sports Analytics · Computer Vision
Client · Fly-Fut

Automated Soccer Highlight Generation

Problem

Match editors were spending hours per game manually clipping highlights — a workflow that couldn't scale to entire leagues.

Approach

A computer vision pipeline that detects game objects (ball, players, referees), tracks player interactions, and identifies highlight-worthy events. Model output feeds directly into an editing tool for one-click clip export.

Impact
  • 5x reduction in editing time per match
  • Consistent highlight quality across leagues
  • Human editor stays in the loop for final approval
Computer VisionObject DetectionAction Recognition
Security camera monitoring an outdoor area
Security · Computer Vision
Client · Crezer SA

CCTV Human vs Animal Movement Detection

Problem

Existing CCTV motion-alert systems were overwhelming operators with false alarms triggered by wildlife, causing real intrusions to be missed in the noise.

Approach

A classifier that distinguishes humans from animals on live CCTV feeds — added as a lightweight component in front of the existing alerting stack so any camera can be upgraded without rewiring.

Impact
  • 80% reduction in false alarms
  • 80% recall on genuine human intrusions maintained
  • Deployed across existing camera infrastructure
Computer VisionImage ClassificationEdge Deployment
Person using sign language
Accessibility · Computer Vision
Client · EldeS

Sign Language Detection Web App

Problem

The deaf and hard-of-hearing community needed a browser-based tool that could recognize sign language gestures for early education and accessibility use cases.

Approach

A web application built on MediaPipe Hand Detection that recognizes numbers 1–10 and letters V–Y in real time from a webcam feed — no install, no hardware, runs entirely in the browser.

Impact
  • Helped the client secure follow-on investment
  • In-browser recognition — no install required
  • Foundation for expanding the recognized gesture set
MediaPipe Hand DetectionWebRTCReal-time inference
Start a Project

Bring us the problem your team keeps avoiding.

A trading system that keeps getting kicked down the road. A model stuck in staging. An audit that's coming. Tell us — you'll get a plan, not a pitch.

Contact usinfo@omextechnologies.com