Aum Kendapadi

Aum Kendapadi

UNC Chapel Hill

Hi! I am an undergraduate student at the University of North Carolina at Chapel Hill, pursuing a B.S. in Computer Science.

My research interests lie in the field of natural language processing (NLP), particularly in the use of large language models (LLMs) and their pedagogical applications. Currently, I am working as a research assistant in UNC NLP’s Learning from Language Lab, where I am advised by Shashank Srivastava. I am passionate about exploring how LLMs can enhance educational experiences and improve teaching methodologies.

In addition to my research, I have a strong background in web development and entrepreneurship. I co-founded and developed Classmate, a social course scheduling platform used by more than 10,000 students at over 10 schools. I have also created other web applications that reflect my love for my community, such as uncGuessr, and my research interests, such as Waffle. I am comfortable working with various full-stack technologies, including React, Angular, Vue, Node.js, Flask, PostgreSQL, MongoDB, Firebase, and Supabase.

Outside of research and web development, I love to play sports, watch movies, and listen to music.

🔥 News

  • 12/24: Check out INTERACT on arXiv
  • 12/24: Reached Platinum in Valorant
  • 09/24: Nominated Phi Beta Kappa (ΦΒΚ)
  • 05/24: Joining UNC NLP’s MURGe Lab as a Research Intern this summer

💼 Experience

  • 07/23 - Now: Learning from Language Lab, Research Assistant
  • 08/24 - 12/24: COMP 311: Computer Organization, Teaching Assistant
  • 05/23 - 11/24: HackNC, Director of Logistics
  • 06/24 - 9/24: MURGe Lab, Research Intern
  • 6/24 - 7/24: COMP 210: Data Structures and Analysis, Teaching Assistant
  • 06/23 - 07/23: Launch Chapel Hill, S23 Cohort

📝 Publications

  • ACL 2025 Submission INTERACT: Enabling Interactive, Question-Driven Learning in Large Language Models
    • Aum Kendapadi, Kerem Zaman, Rakesh Menon, Shashank Srivastava
    • This paper explores how LLMs can transition to interactive, question-driven learning through student-teacher dialogues. We introduce INTERACT (INTERactive Learning for Adaptive Concept Transfer), a framework in which a “student” LLM engages a “teacher” LLM through iterative inquiries to acquire knowledge across 1,347 contexts, including song lyrics, news articles, movie plots, academic papers, and images. Our experiments show that across a wide range of scenarios and LLM architectures, interactive learning consistently enhances performance, achieving up to a 25% improvement, with ‘cold-start’ student models matching static learning baselines in as few as five dialogue turns. (arXiv)

👨‍💻 Projects

sym

Classmate

  • Schedule classes with friends
  • Helping over 10k students at over 10 schools
  • Built with React, Firebase, Mixpanel
sym

Waffle

  • Chat with YouTube videos
  • PennApps 2023 w/ Arvindh Manian, Billy Pan
  • Built with LangChain, Whisper (OpenAI), FastAPI
sym

uncGuessr

  • Guess where photos were taken at UNC
  • Built with Angular, MongoDB, Express.js
sym

Tunetones

  • Create album cover color palettes
  • Built with Vue
  • Spotifind: Generate ML music recommendations - Vectorize feature sets of playlists created using TF-IDF and OHE then find similar songs using the KNN algorithm - Built with Scikit-Learn
  • Twitiment: Analyze the sentiment of tweets using NLP - Analyze the sentiment of tweets using VADER and RoBERTa NLP models, as well a DistilBERT Hugging Face Pipeline - Built with NLTK, Hugging Face
  • Wikipath: Find (the shortest) paths between Wikipedia articles using bidirectional search - Built with Flask, PyVis
  • Digitect: Teach a computer to read handwriting - Use the MNIST dataset to train a homemade neural network to recognize digits - Built with NumPy
  • Fractastic: Create Markovian Lindenmayer systems - Combine L-systems and Markov chains to produce stochastic fractals - Built with p5.js
  • Flowscape: Simulate fluid movement with Perlin noise on vector fields - Built with p5.js
  • Enigma: A basic model of an Enigma machine with plugboard, starting position, and ring functionality
  • Vigenere: A Vigenere cipher with two decryption tools: a brute-force dictionary attack, and a kasiski examination which uses n-gram character repetitions in large text samples to conduct frequency analysis

📚 Education

  • Spring ‘25: University of North Carolina at Chapel Hill
    • B.S. Computer Science
    • Coursework (In Progress): Data Structures and Analysis, Systems Fundamentals, Foundations of (Object-Oriented) Programming, Discrete Mathematics, Models of Languages and Computation, Algorithms and Analysis, Multivariable Calculus, Linear Algebra, Probability, Foundations of Software Engineering, Compilers, Video Recognition, Deep Learning, Files and Databases, Practical Web Design and Development
    • GPA: 3.9
  • Summer ‘24: University of Copenhagen
    • Study Abroad
    • Coursework: Computer Organization (and Architecture)
  • Spring ‘22: Providence High School
    • High School Diploma
    • Dual Enrolled at Central Piedmont Community College
    • GPA: 4.7