Visually impaired individuals and automated cash-handling systems need instant, reliable identification of currency notes in real-time video streams.
Fine-tuned the EfficientNet architecture on a curated dataset of Pakistani notes. Integrated with OpenCV for live video processing and real-time inference.
Related case studies
97–98%
Agentic RAG legal assistant for Saudi regulatory law. 97–98% recall@10 across a bilingual English-Arabic corpus validated by legal domain experts.
300+ GB
Intelligent search system processing 300+ GB of unstructured enterprise data for a major energy company, including 1K–6K page technical documents.
32
AI-powered document intelligence platform with multi-tenant architecture, RBAC, and hybrid retrieval.
Previous
Bilingual Voice Agent on LiveKit
Production-grade real-time voice agent supporting English and Urdu, with sub-200ms response latency and VAD-based natural turn detection.
Next
Urdu Sentiment Analysis — FYP & Publication
Hybrid CNN-RNN model for Urdu sentiment classification, published in a peer-reviewed journal with 96% F1 score.