Yashwant Chouhan

Lead the Implementation of end-to-end machine learning pipeline tailored for a finance client. The orchestration of this solution was streamlined through Azure Data Factory and Azure Databricks. Throughout the pipeline creation, Azure Data Lake Storage and Data Catalog were leveraged for seamless data and report storage. Employed Great Expectations for robust data quality checks, Databricks Feature Store for efficient feature Store, and MLFlow for comprehensive ML experiment tracking, metrics logging, model registration, and parameter logging. Built an automated pipeline for Yield Estimation project. For this project, I utilized Databricks, Azure Data Factory, Azure DevOps, ADLS Gen2 to construct solution pipeline. Additionally, I employed MLFlow for experiment tracking and logging the model. Developed Market Mix Modeling App using flask in python and deployed on google cloud using Cloud Run, Cloud Build for CICD. Utilized BigQuery as the SQL Database. Created automated trigger for refreshing the data without need of app engineering support.

Removed unwanted noise from the images using autoencoder which helped to improve the quality of images, that image also helped to improve the quality of output of OCR. Built solution for passbook analysis problem which gives the statistical data (avg credit amount, avg debit amount, is the passbook holder approved for loan etc.) of the passbook holder. In the passbook analysis, built an object detection model for detecting the rows and columns in the passbook image. Used OCR engine for extracting the text from the passbook images and used rule-based engine for validate the passbook folders. Built the entity detector for front page of passbook which helped to get the information (acc. number, branch name, date of issue etc.) of passbook holder. Developed parsers using regular expressions in python for highlighting the information from the row text data. Developed POC (Proof of Concepts) using MS Azure Services for clients’ problems.

Built pre-processing module for improve the quality of image. Built the solution using object detection algorithm which helped to reduce the time taken by manual process by people and saved the cost for the company. Built assembler to increase the confidence score of the products detected by the various object detection models. Built dashboard for track the performance of the AI models. Built UI using flask and MongoDB where team can see the output of the AI pipeline and check which images rejected by the pipeline. Managed the AI pipeline and improved productivity.

Performed transfer learning on pre-trained model which helped to extract the features from the data. Built Module for localization of the objects present in the images which helped the team to detect the position of object.

Developed website for the NGO which led to increase the donation for NGO. Integrated the payment getaway of third party with the UI. Changed the design of website and hosted on firebase.