Dorcas Kahunguka

I've built backend mobile-money APIs and CRM systems at Meliora Technologies using Java and Spring Boot, then led functional testing and UI/UX review. My design-assumption audit surfaced two critical CRM failures that the client later independently confirmed during on-site testing.

On iLabor, I co-built a non-invasive labor-progress assessment proof of concept, from an Arduino and piezoelectric sensor setup through an end-to-end MySQL data pipeline. The captured time-series data supported a partner's Transformer-based model for cervical-dilation patterns.

I'm implementing RNN, LSTM, and Transformer architectures from scratch through a deliberate theory-first process. I've also built a Python profit-forecasting model and completed GPT-2 probing research that localized typoglycemic failure patterns to tokenization and early attention layers.

I'm developing toward production-oriented ML engineering while continuing independent research into mechanistic interpretability, AI safety, and the origins of machine-learning opaqueness.