Moses Illa

I’m a Machine Learning Engineer and Full-Stack AI Developer with 3+ years building and deploying production AI systems. I focus on computer vision, medical imaging, document AI, NLP, speech synthesis, and financial fraud detection—turning research-grade models into reliable services.

I fine-tune PyTorch models across multiple domains, from DONUT transformer-based invoice extraction to multimodal medical image analysis (X-ray, CT, MRI). I’ve delivered explainability and interpretability using GradCAM and SHAP, so stakeholders can trust what the models highlight.

I manage the full model lifecycle end-to-end: dataset preparation, experiment tracking with MLflow and Weights & Biases, and hyperparameter optimization with Optuna. Then I package and deploy containerised ML services using Docker, AWS/DigitalOcean, and production-ready serving components.

I’m also a solo founder of PrideMatch, a live multi-tenant AI SaaS with real paying customers. I build complete platforms—model, data pipelines, APIs, and monitoring—while designing for security (JWT/RBAC, HIPAA-compliant audit logging, encryption, and DICOM de-identification) and real-world reliability.