Hire Gunashekhar V. - Fine-Tuning Engineer

Developed and fine-tuned a PyTorch-based supervised fraud detection model within BNY's real-time transaction monitoring pipeline, reducing false positives by 23% and improving overall model precision across high-volume daily payment workflows. Engineered end-to-end MLOps pipelines using MLflow, Docker, and AWS SageMaker to automate model retraining and deployment for credit risk scoring, cutting refresh cycles from 14 days to 72 hours. Implemented explainable AI (XAI) frameworks using SHAP and LIME to ensure regulatory-compliant model interpretability, enabling risk, audit, and compliance teams to confidently review all ML-driven credit and lending decisions. Designed and deployed a Retrieval-Augmented Generation (RAG) knowledge assistant using Hugging Face LLMs and Python, integrated with proprietary financial data, reducing institutional banking analyst average research time by 40%. Collaborated cross-functionally with quantitative researchers and software engineers to architect scalable Spark-based feature engineering pipelines on Databricks, powering real-time market risk and portfolio prediction models for asset management division.

Performed feature engineering and data preprocessing for machine learning models using Python and Pandas, transforming healthcare claims data to improve AI model accuracy by 15% through strategic feature selection. Developed machine learning classification models using Scikit-learn and supervised learning algorithms to predict claim approval likelihood, achieving 82% accuracy and streamlining underwriting workflows for processing teams. Created AI-powered member risk assessment system analyzing patient demographics and medical history using Python, Pandas, and regression models, enabling proactive care management for 8,000+ high-risk insurance members. Implemented predictive analytics model using machine learning algorithms and feature engineering to forecast member healthcare expenses, improving premium pricing accuracy by 20% for actuarial teams. Deployed machine learning inference APIs using FastAPI and Docker on AWS EC2, serving real-time AI predictions for claims processing with Git version control, achieving 99% uptime.