Bha Vyak Yadar

Architected enterprise-grade Retrieval-Augmented Generation (RAG) engines using LangChain, LlamaIndex, and vector databases (including Pincone and Chromatoprocessing for complex clinical data). Reduced query response time by 42%. Fine-tuned transformer-based models (BERT, GPT variants) with PyTorch/Hugging Face and AWS SageMaker to improve automated medical text classification and entity extraction by 28%. Built scalable, low-latency model-serving APIs with FastAPI/Flask, containerized with Docker, and orchestrated on Kubernetes on AWS to handle 150,000+ daily requests with 99.9% uptime. Spearheaded end-to-end MLOps automation with MLflow for experiment tracking and real-time inference drift monitoring, cutting deployment cycle time by 50%. Optimized large-scale data pipelines with Apache Spark and SQL to process 5TB+ of health records, increasing extraction speed by 35% for US insurance operations.

Engineered production-ready computer vision pipelines using OpenCV, Vision Transformers, and custom CNN architectures to automate visual inspection workflows and increase image classification accuracy by 31%. Developed custom OCR and document intelligence solutions using deep learning to extract structured data from unstructured enterprise documents, reducing manual data entry overhead by 45%. Built scalable predictive modeling and time-series forecasting solutions using XGBoost, LightGBM, and scikit-learn on large-scale SQL and MongoDB datasets to support revenue optimization. Integrated model explainability frameworks with SHAP and LIME to improve transparency and regulatory compliance for US banking and financial clients. Designed and optimized automated ETL pipelines using pandas, NumPy, and Apache Spark, accelerating EDA cycles by 35% for cross-functional analytics teams. Deployed containerized ML microservices using FastAPI, Flask, and Docker on AWS EC2 and S3, achieving a 30% re

Engineered production-ready computer vision pipelines using OpenCV, Vision Transformers, and custom CNN architectures to automate visual inspection workflows and increase image classification accuracy by 31% for global enterprise clients. Developed custom OCR and document intelligence solutions using deep learning to extract structured data from unstructured documents, reducing manual data entry overhead by 45%. Built scalable predictive modeling and time-series forecasting solutions using XGBoost, LightGBM, and scikit-learn on large-scale SQL and MongoDB datasets to drive client revenue optimization. Integrated model explainability using SHAP and LIME for transparency and regulatory compliance for US banking and financial clients. Designed and optimized automated ETL pipelines using pandas, NumPy, and Apache Spark, accelerating exploratory data analysis cycles by 35% for cross-functional analytics teams. Deployed containerized ML microservices using FastAPI, Flask, and Docker on AWS E