Hemanth Kongara

AI/ Machine Learning Engineer at Charles Schwab

Developed scalable fraud detection pipelines using Apache Spark, PyTorch, and Kafka to reduce detection latency by 60% and improve precision and recall by 25%. Deployed and automated ML models with Azure ML, MLflow, Docker, and Kubernetes, achieving a 40% reduction in deployment cycles and 99% uptime. Engineered deep learning and NLP models for anomaly detection, boosting accuracy by 30% and reducing false negatives by 22%. Applied Bayesian optimization and interpretability tools like SHAP and LIME to enhance predictive performance and automate retraining, improving accuracy by 15%. Incorporated generative AI and retrieval-augmented generation pipelines for fraud analytics, reducing false positives by 18% and increasing coverage by 25%.

Machine Learning Engineer at Seven Fincorp Group

Built document classification systems with Logistic Regression, Random Forest, and SVM to automate document indexing, cutting manual effort by 70%. Designed end-to-end ML pipelines with Python, FastAPI, OpenCV, Tesseract, Hive, Spark, and Kafka to triple document processing speed and reduce latency by 55%. Developed data integration frameworks to improve data consistency by 40% and reduce errors by 30%, enhancing fraud monitoring. Applied NLP for semantic data linking, improving retrieval precision by 35% and risk assessment speed by 50%. Indexed data into Apache Solr and relational databases, boosting customer risk profiling accuracy by 25% and reducing quote turnaround time by 20%.

AI/ Machine Learning Engineer at Charles Schwab

Built scalable end-to-end fraud detection pipelines using Apache Spark, PyTorch, and Kafka, reducing detection latency by 60% and improving precision/recall by 25%. Designed, deployed, and automated production-ready ML models with Azure ML, MLflow, Docker, and Kubernetes, streamlining CI/CD and reducing deployment cycles by 40%. Engineered deep learning and NLP models for anomaly detection, enhancing accuracy by 30% and reducing false negatives by 22%. Applied Bayesian Optimization and explainability tools to improve model performance, boosting accuracy by 15% and cutting monitoring overhead by 20%. Incorporated Generative AI techniques with LLM-powered insights and RAG pipelines to analyze transaction narratives, reducing false positives by 18% and improving contextual detection coverage by 25%.