divinedebelu

Developed a deep reinforcement learning agent for dynamic spectrum allocation in 5G edge networks; results published in IEEE Communications Letters (2022). Curated and annotated a 250K-sample clinical text classification dataset; achieved 91% F1-score with a fine-tuned RoBERTa model deployed in a HIPAA-compliant pipeline. Designed secure data-ingestion pipelines with end-to-end encryption and role-based access controls, aligning with NIST Cybersecurity Framework guidelines. Implemented algorithmic backtesting frameworks for ML-driven trading strategies, integrating financial time-series forecasting models (LSTM, Temporal Fusion Transformer).

Built a sentiment-analysis microservice (Python / Flask) integrated into a customer-support platform, improving ticket-routing accuracy by 34%. Contributed 8 production features across Agile sprints, applying secure SDLC practices including code reviews, static analysis (SonarQube), and dependency vulnerability scanning. Developed a prototype personal-finance dashboard with transaction categorisation, anomaly flagging, and spend-trend visualisation using React and D3.js.