Santhoshkumar Reddy A

About Santhoshkumar Reddy A

I build machine learning systems that make it into production and stay there.Five years designing end-to-end ML pipelines across fintech and healthcare: from feature engineering and model training to AWS deployment, real-time dashboards, and GenAI automation. My work is used daily by credit analysts, underwriters, and clinical teams to drive decisions that matter.What I specialize in: Credit risk & default prediction - XGBoost, Logistic Regression, SHAP explainability GenAI workflows - LangChain, OpenAI GPT-4, RAG pipelines, prompt engineering NLP & document intelligence -SpaCy, BERT, OCR, named entity recognition MLOps infrastructure - Docker, Kubernetes, AWS SageMaker, EC2, S3, MLflow, CI/CD Clinical prediction systems - survival analysis, A/B testing, patient risk segmentationSelected results: 18% lift in 90-day credit default prediction accuracy - Principal Financial Group 60% reduction in executive report prep time via GPT-4 + LangChain automation 22% improvement in OCR field-level accuracy across healthcare enrollment forms - Optum 40% BLEU score improvement on video caption generation (BLIP fine-tuning)Stack: Python · SQL · PyTorch · TensorFlow · XGBoost · LangChain · OpenAI API · AWS (EC2, S3, SageMaker, Lambda, Glue) · Docker · Kubernetes · MLflow · Spark · SHAP · SpaCy · BERT · FLAN-T5 · YOLOv5 · Tableau

Experience

1. Ai Ml Engineer

   Principal Financial Group

   Jan 2025 — Present · US

   CreditWatch AI - GenAI-Powered Risk & Portfolio Intelligence Platform Designed and deployed a credit default risk classification system (XGBoost, Logistic Regression) trained on 70K+ account records -achieving an 18% improvement in 90-day default prediction accuracy over legacy bureau scorecards. Engineered 45+ behavioral and transactional features using SQL and Pandas quantifying repayment patterns, credit utilization, and delinquency streaks - driving an 11% AUC gain through feature enrichment alone. Built a Streamlit internal dashboard with SHAP explainability, lift charts, and confidence bands adopted by credit analysts and underwriters for weekly portfolio risk reviews. Architected a GPT-4 + LangChain pipeline to auto-generate executive portfolio summaries from statistical anomaly flags - cutting report preparation time by 60% and improving decision clarity for senior leadership. Containerized the ML pipeline with Docker and deployed on AWS EC2 with scheduled batch inference; outputs piped to S3 and surfaced via Tableau dashboards used in monthly risk reviews. Partnered with risk strategy, credit ops, and compliance to validate model thresholds and align outputs with FCRA and FDIC documentation requirements.Skills: XGBoost · LangChain · GPT-4 · SHAP · Streamlit · Docker · AWS EC2/S3 · Tableau · Python · SQL · FCRA Compliance

Education

• GITAM Deemed University

  Bachelor of Technology - BTech, Computer Science

• University of Wisconsin-Milwaukee

  Master's degree, Data science