At Technical University of Berlin, I deployed an XGBoost predictive ML pipeline on an HPC cluster to forecast German spot prices. The resulting ML-guided decision framework reduced hedging costs by 18% compared with rule-based baselines.
I also analyzed terabyte-scale distributed BGP routing telemetry on HPC systems, identifying stability patterns that informed high-throughput compression algorithm design. At PricewaterhouseCoopers (PwC), I partnered with client stakeholders on enterprise rollouts, developing ETL pipelines and backend APIs for financial and inventory data.
In my projects, I built a computer vision pipeline to detect honeybee waggle-dance behaviors and used distributed GPU training to reduce per-epoch training time from three hours to 20 minutes. I’ve also built a self-healing Linux infrastructure agent and a multilingual RAG chatbot.
