Yashwanth Reddy

Engineered and trained Python-based ML models for trading and risk workflows, owning feature development, validation, and production handoff. Delivered models reaching ~94% precision on evaluation sets. Built and maintained an LLM-based RAG workflow using vector search, prompt design, grounding, and response evaluation to support trading and risk analysts. Architected and deployed containerized ML services on AWS using Docker and Kubernetes with MLflow for experiment tracking and controlled promotion to production. Implemented automated data-quality and drift monitoring with retraining/rollback, reducing unplanned retraining incidents by 25% and improving production stability. Optimized Python/SQL ETL pipelines for near-real-time inference, reducing data-processing latency by 30%, and designed release checks using A/B testing and regression validation to catch performance regressions earlier. Mentored junior engineers through code/design reviews and debugging, partnering with risk, dat

Developed predictive and anomaly-detection models in Python for healthcare transaction data, including data profiling, feature engineering, model comparison, threshold tuning, and validation over 500K+ records. Built and optimized BERT/Transformer NLP pipelines to extract and classify information from clinical documentation, reducing repetitive manual review time by 35%. Used AWS SageMaker for deep-learning training and built reusable validation checks for schema changes, missing values, and distribution shifts before retraining. Deployed ML functionality via FastAPI and AWS Lambda with an emphasis on response-time requirements, maintainability, and observability/cost. Tuned decision thresholds to improve discrimination while controlling false positives in operational workflows, and documented assumptions, data lineage, validation methodology, deployment steps, and runbooks to support governance and audit requirements.

Delivered healthcare ML solutions across the lifecycle: exploratory analysis, feature engineering, training, validation, and integration with application services; achieved an 18% diagnostic accuracy improvement on the project evaluation set. Built SQL and Apache Spark ETL pipelines for large structured datasets used in training and batch scoring, adding completeness/schema and failed-record checks to improve data reliability. Created FastAPI inference services and collaborated on API contracts, error handling, logging, integration testing, and production support for low-latency predictions. Optimized distributed NLP data-preparation workflows and reusable components, reducing preprocessing repetition and accelerating pipeline runtime by 40%. Automated build/deployment through CI/CD and contributed to unit/integration testing, code reviews, sprint planning, and production releases. Applied hyperparameter tuning and 5-fold cross-validation for model selection and overfitting diagnosis.