Apidhe Osii

At TushindeBet, I built fraud detection and reconciliation systems processing about 1.5 million transactions monthly with 99.95% uptime and zero data loss. My work cut missing-funds customer care calls by 50% and reduced average case resolution from roughly eight minutes to under 90 seconds.

I added an XGBoost risk model and human-in-the-loop retraining workflow that keeps false positives below 2% while reducing manual review volume. I also build real-time Node.js services, WebSocket systems supporting 2,500+ concurrent users, and operational dashboards that shortened incident detection to under two minutes.

I'm currently building Bongo, an adaptive learning platform that converts printed past papers into personalized practice. I've processed 500+ scanned papers into 12,000+ structured questions and built a LangGraph pipeline that generates 50–80 targeted questions per learner session.

Across Lottomotto and TripleP TV, I've delivered financial transaction services, event-driven pipelines, REST APIs, automation workflows, and dashboards that replaced slow manual reporting. I care about systems that work reliably first, then improve over time through data, feedback, and practical AI.