I am Sree Dhyuti

Education & Experience

Academic Background

Master's in Artificial Intelligence

Sept 2024 - Dec 2025

Northwestern University, Evanston, IL

CGPA: 4.0/4.0

Focus: NLP, Deep Learning, GenAI, Machine Learning, Graph Neural Networks, Data Science, High Performance Computing

M.Tech + B.Tech in Computer Science Engineering

Jul 2019 - May 2024

Indian Institute of Information Technology, Kancheepuram, Chennai, India

CGPA: 8.86/10.00

Focus: Big Data Analytics, Digital Image Processing, Computer Graphics, Pattern Recognition, Computer Vision

  • Presented final year thesis on "Improved Text-Summarization with PEGASUS and Siamese Network Evaluation" at IEEE TENCON 2024 (Singapore)
  • Ranked 37th globally out of 1,500 teams in the OpenCV AI Competition 2022

Professional Experience

AI Research Intern

Jun 2025 - Sept 2025

Writer, San Francisco, CA

AI Researcher

Mar 2025 - Jun 2025

The Abazeed Lab, Chicago, IL

ML Research Intern

May 2023 - Oct 2023

BioSystems & Controls Lab, Chennai

ML Intern

Aug 2022 - Dec 2022

Tiny Banyan Technologies, Chennai

Portfolio

A not-very exhaustive collection of all of my projects

  • All Projects
  • Artificial Intelligence
  • Machine Learning
  • Data Science
  • Computer Vision
  • NLP
  • General

REINFORCE-based Contradiction Correction in LLMs

Built a BERT-based contradiction detection model fine-tuned with LoRA and improved by a REINFORCE RL framework, achieving ~90% detection accuracy and reducing contradiction rate from 65% to 44%.

Dialogue Summarizer

Fine-tuned PEGASUS LLM on SAMSUM corpus and deployed a FastAPI- and Docker-based summarization app with AWS and CI/CD integration, streamlining workflows and automated testing.

AI Powered Hotel Recommendation System

Enhanced natural language understanding in GPT-4o by fine-tuning retrieval mechanisms using agents and RAGs to reduce hallucinations, delivering real-time, preference-aligned hotel recommendations with higher precision.

Drugs, Side Effects, and Medical Conditions Analysis

This project involves analyzing a pharmaceutical dataset to explore the relationships between drugs, side effects, and medical conditions using data visualization, association mining, and various other techniques.

Human Artery Vein Classification with U-Nets

A U-Net based deep learning model built to classify and differentiate between human artery and vein images for medical diagnostics and research.

Bot Clean

A simple bot designed to navigate through a matrix and clean all of its dirty cells.

QnA Bot

A TF-IDF based parser to parse through paragraphs and answer questions.

Color-based Classification

A Computer Vision Program to detect the color of the balls and get the timestamps of each ball entering and exiting the quadrants.

PHP-based Student Management Portal

A PHP-based student portal that helps students manage their academic and extracurricular activities, featuring modules for login, timetable, lost and found, assignments organizer, password change, and internship updates.

CS50 AI Mini-Projects Repository

A collection of 12 unique AI programs developed using various AI concepts, including Optimization, Machine Learning, and NLP.

Social Distancing Monitor

An image processing project that utilizes an inbuilt Histogram of Oriented Gradients (HOG) descriptor to recognize people in a given area and estimate the distance between them

CNN-based Leaf Disease Classification

A CNN-based classifier with image segmentation to accurately detect and classify plant leaf diseases using the Plant-Village dataset.

Diffusion Modeling from scratch

Implementation of a Conditional Denoising Diffusion Probabilistic Model (DDPM) built from scratch for image generation tasks.

Blog

My writings on AI

Digital Image Processing: A Primer (Part 1)

Published on Medium

An introduction to the fundamentals of digital image processing, covering core concepts and techniques used in computer vision applications.

Read Article

Neural Networks That Learn Without Activation Functions

Published on Medium

Exploring alternative neural network architectures that challenge conventional wisdom by learning effectively without traditional activation functions.

Read Article

Music

I love music and find myself immersed in melodies as I breathe. I sing and create music on FL Studio :))

Listen to my music on SoundCloud:

Contact

Address

Ridge Ave, Evanston, IL - 60201

LinkedIn

Sree Dhyuti Nimmagadda

Email

sreedhyutin@gmail.com