Rohith Thodime

Spearheaded a team of 3 specialists in a 1,600-hour initiative to engineer and deploy a pioneering Agentic AI framework, revolutionizing the firm's Optical Character Recognition (OCR) capabilities. Architected a sophisticated multi-agent system on AWS that automates data extraction, validation, and correction.

Successfully led the team and implemented a 4,400 hour project to build out two Policy Engines for the SNAP & TANF policy teams for one of the states of USA. This solution had 95% accuracy, it was built using entirely managed services in AWS, making it Serverless and minimizing the overall ongoing maintenance cost to the client.

Developed Machine Learning models for Writer & Non-Writer for JPH in the Dataiku Pipeline and created the Scenarios through the Dataiku pipelines. Currently working over Personalized Medicine, using Generative AI – Prompt Engineering & Fine Tuning LLMs, to help match patients with the most suitable treatments.

Focused on AIOps and Cognitive Robotic Process Automation that leverage Artificial Intelligence. Used Amazon Lex to build the Conversational AI chatbot, thereby adapting Intelligent Process Automation with seamless integration of Intelligent Cognitive Conversational Chatbot.

Built Machine Learning models in PySpark in AWS Sagemaker and developed Scorecards from Probability values generated from Logistic Regression Model. Identified important features using AIC, p-value, and Random Forest, selecting the best model based on AUC and F1 score.

Focused on building and testing Automated Machine Learning and AI models in PySpark in the backend, then integrating them into the Business Intelligence product. Developed models to trigger Classification or Regression ML models based on data type, or unsupervised ML models if no target variable.

Worked on predicting the Defaulter’s Probability of Payment of Outstanding Balance (Propensity Score) to create a 3x3 Matrix of Propensity score vs Outstanding Balance, defining strategies for the Tele-Calling team. Developed Machine Learning models like Logistic Regression, Decision Trees, Random Forests, ADA Boost, and GBM to predict the Right Time to call a Defaulter.

Analyzed geospatial data of over 2 crore Dwellings, Houses, Landmarks, and Road Networks across India. Generated valuable insights and visualizations from real-time streaming data of Apache Kafka in MongoDB using Python.

Worked with the Research Team at Novartis HQ in Switzerland to develop Machine Learning models for Clinical Trials. Gained practical experience in applying data science techniques within a pharmaceutical research environment.

Graduated from the Advanced Management Program in Business Analytics from the Indian School of Business (ISB). This program focused on enhancing business analytics skills.