Jason Mloza@jason_mloza
1
Events attended
1
Submissions made
Malawi
1 year of experience
About me
I am an aspiring Earth Science student at Malawi University of Science and Technology (MUST) with a strong interest in applying artificial intelligence to real-world industrial challenges. My focus lies at the intersection of geology, data science, and emerging technologies, where I aim to develop intelligent systems that improve decision-making in sectors such as mining and resource management. I am currently building GeoMind AI, a project that uses machine learning to analyze geological data and predict high-potential mining zones with clear, explainable insights. This reflects my broader goal of creating practical AI solutions that are not only technically sound but also economically impactful. Beyond my academic path, I actively explore fields such as robotics, software development, and advanced computing, with a long-term vision of building scalable technologies that can operate across global industries. I am particularly interested in leveraging high-performance computing and GPU acceleration to solve complex scientific problems. Through the AMD AI Developer Program, I aim to deepen my understanding of AI infrastructure, optimize models for performance, and collaborate within a community pushing the boundaries of applied intelligence
🤓 Latest Submissions
MineralIQ - Gold Exploration AI
MineralIQ is an explainable geospatial AI system that transforms multispectral satellite imagery into actionable gold exploration intelligence while reducing early-stage fieldwork. Users click any location on an interactive global map. MineralIQ queries Google Earth Engine for Sentinel-2 Surface Reflectance imagery (2023 median composite, cloud filtered) and performs 8-band spectral analysis across a 20 km region of interest. The system extracts geological indicators associated with gold mineralisation: Iron Oxide (B4/B2), Clay Mineral Index (B11/B8), NDVI, SAVI, RVI, NDII, MGI, and Thermal SWIR Ratio. These features are normalised and fused into a single anomaly score visualised as a heatmap where green indicates low, blue medium, and red high mineral potential. MineralIQ is trained using known gold deposit coordinates and surrounding geological terrain patterns, learning relationships between spectral signatures, mineral alteration zones, and land formation characteristics. A complementary XGBoost model uses elevation, slope, and distance-to-deposit data to improve prediction confidence. This probabilistic targeting significantly reduces exploration costs by allowing companies to prioritise only high-probability zones, reducing unnecessary drilling, field surveys, and land disturbance. This lowers both operational costs and environmental damage. MineralIQ demonstrates how mineral exploration can shift from expensive, invasive surveying to intelligent satellite-driven targeting. This is currently a focused prototype trained on a limited gold dataset to validate the methodology. Future versions will scale using larger geological datasets and AMD GPU accelerated computing for faster raster processing and improved model accuracy. Future expansion includes copper, lithium, and rare earth mineral detection, evolving MineralIQ into a global mineral intelligence platform.
10 May 2026
