Integrating Computer Vision and Predictive Models into mobile commerce.
Extensive experience creating high-quality training datasets for foundational models. Collaborated with industry leaders CrowdFlower, BasicAI, and Samasource to engineer robust datasets for Computer Vision and NLP model alignment.
Real-time mesh mapping on iOS using ARKit/RealityKit. Allows users to virtually try on sunglasses and hats with 98% facial landmark accuracy.
Edge-based visual search for retail staff. Identifies 500+ SKUs instantly without internet connection, optimized for low-end Android devices.
A purchasing assistant that understands local dialects (Swahili/Sheng mix). Analyzing intent...
Machine learning model analyzing M-Pesa transaction patterns to predict stockouts for 200+ retail locations with 85% accuracy.
Advanced sentiment analysis system processing customer reviews, social media posts, and feedback in real-time. Classifies emotions (positive, negative, neutral) with 92% accuracy across multiple languages including English, Swahili, and local dialects. Powers customer experience insights for e-commerce platforms.
Convolutional Neural Network trained to classify product images across 50+ categories. Deployed for automated inventory management, reducing manual cataloging time by 70%. Achieves 94% precision on fashion and electronics datasets.
Hybrid recommendation engine combining collaborative filtering and content-based approaches. Increases average order value by 35% through intelligent cross-selling and upselling. Processes user behavior patterns from 100K+ transactions monthly.
Transformer-based model generating concise summaries of product descriptions, reviews, and articles. Reduces content length by 80% while preserving key information. Integrated into mobile apps for faster content consumption and improved user engagement.
Real-time anomaly detection system identifying fraudulent transactions and suspicious user behavior. Uses ensemble methods combining Isolation Forest and Autoencoders to achieve 96% detection rate with less than 2% false positives. Protects mobile payment systems from financial fraud.
Neural machine translation service supporting 15+ African languages including Swahili, Kikuyu, and Luo. Powers multilingual e-commerce platforms, enabling seamless communication across diverse markets. Achieves BLEU scores above 40 for major language pairs.
Optical Character Recognition system extracting text from receipts, invoices, and ID documents. Handles handwritten and printed text with 89% accuracy. Automates data entry processes for retail inventory management and customer onboarding workflows.