Dhruv Sureshkumar Patel

· Elevated customer service for a client by engineering a system that daily transcribes 8000+ calls using Microsoft Speech to Text, performing linguistic analysis, and concurrently achieving a 25% reduction in transcription costs

· Crafted and executed a comprehensive Data and MLOps Software Development Life Cycle (SDLC) strategy on Azure Databricks and Azure DevOps for enterprise-wide implementation

· Architected a production-ready data mart engine in Snowflake by writing stored procedures using Snowflake Scripting which seamlessly orchestrated table creation, data loading, data masking, and comprehensive logging functionality

· Orchestrated the migration of over 20TB of data from on-premise to Azure by developing data ingestion, purging and validation pipelines within Azure Data Factory

· Enhanced efficiency of large-scale data pipelines with PySpark using Azure Databricks, reducing Azure compute costs by 60%

· Attained 96% precision/recall using Azure Custom Vision API for document classification and employed Azure Document Intelligence API for accurate text, key-value pairs, and table data extraction

· Classified mortgage document images into 300+ classes with a CNN model in Keras achieving accuracy of 95%

· Utilized OpenCV library for image smoothing, morphological transformations, line removal, and contour detection

· Developed production level Python scripts for extraction of texts from documents attaining accuracy of 93.5%

· Trained GPU based open-source OCR engine, Calamari, on 15,000 image segments using transfer learning from TensorFlow checkpoints obtaining a character error rate of 0.10

· Presented classification-extraction workflow in 3 POC (Proof of Concept) sessions to multiple clients

· Engaged in functional testing, A/B testing, and quality improvements processes to provide robust and intuitive functionalities