At Imperial ESE, I architected machine learning pipelines for computational simulation data relevant to asteroid disruption and Martian moon formation. I developed predictive SPH frameworks to process billions of data points on the Imperial CX3 HPC cluster.
At Dans Multi Pro, I designed a real-time fraud detection system, benchmarking anomaly detection models and implementing containerized FastAPI services. The ML architecture I designed increased precision by 15% and recall by 12% over baseline systems.
At LinkAja, I developed and deployed four ML models on GCP for fraud detection, preventing 95% of monthly financial losses and reducing fraud incidents by 20.3%. Earlier, at LAPAN, I applied machine learning and data analytics to satellite-based atmospheric data and co-authored two publications.
