Ankur Chaudhary

Worked on Classifying User based transaction project. Implemented the Neo4j Graph Data Science model on an Azure sandbox instance. Designed and implemented a ML pipeline to ingest a graph data model, compute graph data science features, and train a machine learning model for node classification based on property values. Implemented REST services for online ML training and prediction using trained weights, embeddings, and clustering techniques. Integrated Kafka to handle message drops in Kafka topics, allowing for failure scenario replay. Utilized Python and the Flask framework with Fast API to develop modules and expose URLs on Swagger for easy API access and documentation.

Analysed the project, Database Mapping using NLP, and built a suggestions recommender system utilising NLP and Sentence Transformers. Worked on the Finance Image Classifier Accelerator project for Synechron Innovation Lab. Developed and deployed a ML backend application as a Docker image on an AWS cloud EC2 instance, ensuring efficient deployment and scalability. Converted R scripted modules into Python programming language for seamless integration and improved efficiency. Utilized the Dataiku tool to integrate the Morgan Pipeline, enabling streamlined data processing and analysis. Designed and implemented a machine learning workflow using time series modeling techniques on the GCP platform. Developed build and deploy scripts using Python to automate the deployment process and ensure efficient execution.

Responsible for supporting employee development through training initiatives. Assessed, planned, developed, and delivered various training programs. In the Customer Churn Prediction project, analyzed the rising customer churn rate and declining Average Revenue per Unit (ARPU) for a telecom company operating in a competitive market environment.