Udayan Pandey

Experience

• 

  HCL Technologies

  May 2022 - Jul 2022

  Software Engineer Intern

  • Worked with industry professionals remotely to help develop real-time information-based uses of IoT• Used Python to help implement IoT solutions with real-time information based on new data models for clients• Implemented improved IoT based-services for the local office

• 

  The Data Mine - Purdue University

  Aug 2022 - May 2023

  Undergraduate Data Science Researcher

  • Helping Inogen improve their services with the help of machine learning and data analysis. • Originally, focused on using the company provided data to detect and analyze common patterns, then using these samples to suggest meaningful improvements for their service.• Later on, pivoted to providing the company with an OCR based PDF reading technology to help them understand their records, working with automation software UiPath to create an automated data extractor.• Finished with creating a PDF reader based in Python, using Tesseract and OpenCV2, which converted PDF documents to text with high accuracy rate Show less

• 

  Purdue University

  Sept 2022 - Aug 2023

  • Working under Professor Yung-Hsiang Lu on the ongoing study of Conversations with ChatGPT and Copilot about C Programming • Collaborating with a team to ask ChatGPT and Copilot advanced questions in the C programming language, based on the Purdue course ECE 264• Comparing the results provided by the two models with exam solutions to evaluate their accuracy in handling advanced programming concepts• Analyzing the results obtained from ChatGPT and Copilot to assess the strengths and weaknesses of the two LLMs with regards to the objectives of ECE264• Based on the findings, suggest meaningful changes to the course to reduce student abuse of the LLM models and facilitate a better integration of them with the regards to the course objectives Show less

  • Undergraduate Research Assistant

    May 2023 - Aug 2023

  • Academic Tutor

    Sept 2022 - Aug 2023
• 

  Indian Institute of Technology, Kanpur

  Jul 2023 - Dec 2023

  Data Science Intern

  • Fine-tuned Meta's LLM Llama-2 on Indian epics, Ramayana and Mahabharata, using Pytorch, Langchain, and HuggingFace• Preprocessed data with regex to remove extraneous details, and created a custom dataset with the 7-B Llama-2 model's tokenizer and HuggingFace• Employed Quantization techniques like LORA and QLORA to reduce the model to 4-bits, enabling it to run on Google Colab’s 16 GB GPU.

• 

  The Data Mine - Purdue University

  Aug 2023 - May 2024

  Student Project Lead

  • Mentored two separate teams, Webee and HUMN Capital, consisting of 30+ students functioning as a Scrum Master within an Agile framework, serving as a liaison between the Corporate Partners and the students, resulting in streamlined project execution• Directed the Webee team in developing an image segmentation computer vision model for geospatial feature extraction from satellite imagery• Led the HUMN Capital team in building a full-stack NLP sentiment analysis tool to evaluate emotion swings in manager-employee meetings Show less

• 

  Purdue University

  Jan 2024 - now

  Undergraduate Researcher @ Kihara Lab

  • Collaborating with Professor Daisuke Kihara at Purdue University to design a web server for the GO2SUM Large Language Model, featured in the upcoming published book chapter “Bringing a GO term list to human readable function description using GO2Sum”• Integrated machine learning models into the backend using Python-based RESTful APIs, reducing processing time by 10%• Upgraded outdated Perl and PHP scripts for SQL queries, improving server efficiency and ensuring compatibility with the updated database infrastructure Show less

• 

  Purdue Vertically Integrated Projects (VIP)

  Jan 2024 - Apr 2024

  Undergraduate Researcher

  • Led a team to work on object classification for the application of identifying pedestrian patterns during Covid-19 lockdown using deep-learning framework TensorFlow, presenting a poster at the Purdue Spring 2024 research conference• Adapted and integrated state-of-the-art Recurrent Convolutional Neural Network (RCNN) models for accurate detection and classification of pedestrians• Improved the model by implementing features such as Region Proposal Networks, IoUs and ReLu activation functions to get a final accuracy of 93% on detecting civilians Show less

• 

  Purdue Polytechnic Institute

  May 2024 - Dec 2024

  Software Engineer Intern

  • Engineered a multipurpose NLP-based learning website for 1000+ Purdue students leveraging Django, JavaScript, HTML, and CSS to enable users to create,practice, and analyze PL/SQL questions• Optimized PostgreSQL schema and built a RESTful API with Django, enhancing query performance by 30% and reducing response time by 10%• Integrated Google’s Gemini Generative AI API to dynamically generate sample database tables improving database comprehension by 50%

• 

  The Data Mine - Purdue University

  Jan 2025 - now

  Student Project Lead

  • Led the development of digital QC logbook applications using Tulip and Power BI for AstraZeneca Mt. Vernon team, with goal of replacing paper-based systems and improved workflow efficiency by 20% across supply chain teams• Functioned as a Scrum Master within an Agile framework, serving as a liaison between AstraZeneca and students, resulting in streamlined project executions utilizing tools such as kanban boards• Collaborated with a 14-student team as a part of Purdue’s Data Mine program, aligning development with AstraZeneca’s digital transformation goals through sprint planning, LEAN principles, and cross-functional coordinatio Show less