Adaptive Demand Optimization Engine, Python, LightGBM, AWS Lambda, Designed a dynamic demand prediction engine using LightGBM with automated model refresh cycles triggered via AWS Lambda., Integrated batch inference jobs with serverless execution, cutting infrastructure cost by 38% while maintaining SLA targets., Implemented feature drift monitoring and alerting to preserve stable model performance over time. Graph-Based Customer Journey Analyzer, PyTorch Geometric, Neo4j, FastAPI, Modeled customer navigation and interaction patterns as graphs and trained graph neural networks (GNNs) for multi-step behavior prediction., Improved multi-step churn prediction accuracy by 28% compared to baseline models., Served learned embeddings and insights via a FastAPI service, enabling downstream teams to integrate journey intelligence into tools. Real-Time IoT Stream Conditioner, Kafka, Spark Streaming, Delta Lake, Engineered streaming pipelines to clean, normalize, and persist high-volume sensor data from Kafka into a Delta Lake-backed data lakehouse., Reduced anomaly detection latency by 55% using incremental micro-batch processing and optimized Spark jobs., Built operational monitoring dashboards tracking data freshness, throughput, and pipeline health. LLM-Assisted BI Query Translator, OpenAI API, LangChain, PostgreSQL, Developed an LLM-powered assistant that converts natural language business questions into optimized SQL queries against a PostgreSQL warehouse., Automated approximately 60% of manual query authoring for BI teams, accelerating dashboard and ad-hoc analysis development., Implemented RAG-style context enrichment to align generated queries with business logic and governance rules.
