Gratitude Mbela

Built and leads Hulsalez, a commerce and AI intelligence platform for East African SMEs integrating a fractal AI engine for customer classification, transaction anomaly detection, and fraud prevention.

Designed and implemented a fractal clock–grounded feature representation framework using prime mathematics and lattice theory for ECG time-series classification. Achieved 89% accuracy with Random Forest and 99% with a ConvNet on PTB-XL using limited leads, and is extending the approach with Hidden Markov Models and Factor Graphs.

Received technical mentorship and peer review of computational models and machine learning implementations throughout an independent research program.

Collaborated on time-series data analysis in botanical and ecological research, focusing on gaps in structured data acquisition and applying analytical methods to biological datasets.

Was formally highlighted by Wouter Gevaert for inclusion in an international innovation project through SensiML, which develops embedded machine learning solutions for edge devices.

Applied the fractal pipeline to phoneme-level acoustic modeling to produce stable topic-conditioned clustering for surname-type structures. Demonstrated representational generality across unrelated domains.

Undertook three years of independent study and original research spanning abstract algebra, lattice theory, fractal geometry, signal processing, Bayesian inference, and AI safety, validated on real datasets without institutional support.

Completed foundational computer science coursework before redirecting to self-directed study in advanced mathematics, signal processing, and machine learning theory.