Amey Ghate

Experience

• 

  Tata Motors

  May 2017 - Jul 2017

  Data Analytics Intern at Engineering Research Centre
• 

  Microsoft

  Jun 2017 - May 2018

  Microsoft Student Partner

  Microsoft Student Partner - DA-IICT, Gandhinagar

• 

  IIT Bombay

  May 2018 - Jul 2018

  Research Intern

  Worked as "Ekalavya Scholar" on a platform for Project Based Learning (PBL) using Open Source tool Django.

• 

  DA-IICT Center for Entrepreneurship and Incubation (DCEI)

  Jul 2018 - May 2019

  Vice President – People

  As the Vice-President People DCEI (DA-IICT Centre for Entrepreneurship and Incubation). I interacted with professors, students, entrepreneurs from Gujarat and budding industrialists to analyse and support startup concept

• 

  Cummins Inc.

  May 2019 - Jul 2019

  Intern

  Used Digital Image Processing using Python's OpenCV library to provide safety for manual winding machines. Alerts were provided for potential hazard conditions.

• 

  AlgoAnalytics

  Jan 2020 - Apr 2020

  Data Science Intern

  A Home Equity Line of Credit (HELOC) enables homeowners to leverage the equity in their homes for borrowing. This flexible financial product operates similarly to a credit card, allowing homeowners to withdraw, repay, and borrow again. To ensure responsible lending, companies must assess the creditworthiness of applicants through a rigorous underwriting process, typically managed by underwriting experts.To enhance this process, we embarked on an initiative to automate underwriting using machine learning. We gathered comprehensive credit data, encompassing loan applications, underwriting outcomes, and actual repayment histories. A thorough data analysis was conducted to identify value ranges, missing data, and anomalies. We performed exploratory data analysis on the cleaned dataset to uncover correlations and identify key features relevant to machine learning applications.We evaluated multiple machine learning algorithms to identify the most suitable methods for our dataset, ultimately developing predictive models for HELOC repayment utilizing XGBoost and Feedforward Neural Networks (FFN). Our models achieved an impressive accuracy rate of 89%. Each loan application was categorized as approved, rejected, or requiring manual underwriting. The automation of approved and rejected decisions significantly improved customer response time by 30%, streamlining the lending process.To operationalize this innovation, we developed an automated loan approval recommendation system by deploying our machine learning model as a REST API. The implementation was carried out using Python and Flask, facilitating seamless integration into existing workflows.This project not only enhances the efficiency of the underwriting process but also improves customer experience, demonstrating the power of machine learning in the financial services sector. Show less

• 

  Publicis Sapient

  Oct 2020 - May 2023

  Spearheaded visualization team for Portfolio Analytics' and Portfolio Builder' application Demonstrated projected portfolio returns based on investment strategy and economic factors

  • Associate Software Development Engineer II

    Apr 2022 - May 2023

  • Associate Software Development Engineer I

    Jul 2021 - Mar 2022

  • Associate Technology

    Oct 2020 - Jul 2021
• 

  Microsoft

  Jan 2024 - May 2024

  Data Scientist - Capstone Project

  Microsoft Copilot, an AI-powered assistant integrated into Microsoft 365 applications like Word, Excel, PowerPoint, Outlook, and Teams, leverages advanced AI models, including large language models (LLMs) like GPT, to enhance productivity and streamline user tasks.One use case focused on using Copilot to address support tickets, where the required information is sourced from a knowledge base presented as web (HTML) pages. These pages often include tabular data, which is easy for humans to interpret but challenging for LLMs to understand accurately.The project aimed to significantly improve Copilot’s ability to interpret and respond to questions based on tabular data. I designed and implemented a data engineering ETL pipeline in Python to detect, process, and convert table data into a format comprehensible to Generative AI models. This transformed data is then utilized in Retrieval-Augmented Generation (RAG) processes, resulting in 100% accuracy in responses generated by the AI. Show less

• 

  Smart IT Frame LLC

  Jul 2024 - Dec 2024

  AI Engineer

  Developed an innovative and scalable application leveraging Generative AI to automate manager feedback. Designed and implemented a solution using OpenAI, incorporating advanced prompt engineering and Retrieval-Augmented Generation (RAG), to deliver high-quality, consistent, and personalized employee feedback based on inputs collected over time, while significantly reducing feedback creation time for managers. The generated feedback is displayed to managers through a web portal developed in React, allowing them to review and confirm feedback before communicating it to employees. Show less

• 

  Fractal

  Dec 2024 - now

  Senior Data Scientist

  Engaging in advanced analytics and harnessing the power of GenAI and machine learning, I collaborate with an exceptional team to deliver impactful solutions in the healthcare domain. My work centers on utilizing data to uncover actionable insights that enhance patient outcomes, streamline operational efficiency, and drive innovative strategies across the industry.