Ellora Tapales Devulapally

Computer Science major and English minor skilled in full-stack development and interested in exploring the intersection of education, humanities, and technology.

Skills

Digital Product Management & Shipping

• Worked as an junior product manager on Tapestry’s website components such as coach.com and katespade.com, developing UX/UI workflows for Coach’s product detail page
• Worked in tandem with UX/UI designers, software development team, data analytics team, senior product managers to collaborate on web design.
• Check out my retrospective on my internship here.

Graphic Design

• Proficient in Adobe Software (Illustrator, InDesign, Photoshop, AfterEffects). Read more about my work here.
• Experience in Unity creating visually interesting computer graphics and game design

UX/UI Design & Digital Marketing

• Developed curriculum for Boston University on digital marketing strategy, audience targeting, SEO/SEM, social media, and email marketing.
• Conducted competitive analyses for Delware North, using key tools like Google Analytics to develop strategies for web-based campaigns.

Front-End Development

• Proficient in HTML, CSS, React, TypeScript & JavaScript
• Integrated web apps with front-end interfaces using Flask and Redis

Back-End Development

• Familiar with Git Workflow, Docker and CI/CD
• Experience with database setup in MongoDB and Redis
• Experience using OpenAI APIs, Hugging Face & GPTs; developed key skills in Python & Flask

Education

Davidson College

Bachelor of Science in Computer Science, Minor in English

Expected Graduation 2026

Recent Work

End-to-End ML Prediction Pipeline

I built a full data workflow, from data collection and cleaning to feature engineering and model training, to understand what drives interaction on anonymous social media posts. You can read more about the project here, and check out this website for the applications of our findings.

Tapestry: Web Design & Digital Product Management

At Tapestry, a fashion conglomerate that owns both Coach and Kate Spade, I interned under the senior product manager for the product details page. You can read more about my work here.

Web App Development: Atlas Course Advisor

The Atlas Course Advisor System is a 24/7 AI-powered chatbot designed to assist students at Duke University with course selection, requirements, and resources. This academic advising tool leverages advanced language models and semantic search to enhance the advising experience. Key features include personalized responses, in-chat course details, and deployment on Kubernetes platforms such as OKD and Azure. Sponsored by Duke University, NCShare, and the National Science Foundation (NSF).


View Handout (PDF)
Watch Final Presentation

Zine Design: Side B

I work as a layout editor for my school's alternative music zine. Click here and here to see the most recent mockups.

Coursework Projects

CSC 362 — Data Visualization

Labs and mini-projects exploring storytelling with data (charts, maps, and interactive graphics).

View Coursework Page

CSC 363 — Human Computer Interaction

Homework projects utilizing human-centered design practices.

View Coursework Page

CSC 371 — Machine Learning

Homework projects using linear/logistic regression, KNN, Random Forests, and Neural Nets.

View Coursework Page

Predict the Yik: Modeling Engagement on Davidson's YikYak

Built an end-to-end NLP pipeline to predict post engagement on Davidson's YikYak. Scraped real YikYak and Reddit data, engineered sentiment, linguistic, and temporal features, and compared Logistic Regression, SVM, and Random Forest models under severe class imbalance. Deployed the final model as an interactive Streamlit app.


Read the Paper
Try the App
JIT Compiler Visualization

JIT Compiler Phase-by-Phase Visualization

An independent study with Dr. Katy Williams building an interactive force-directed graph of V8 JIT compiler intermediate representations using JavaScript and D3.js. Lets developers step through each optimization phase to see which nodes are alive, how edges change, and where bugs may have been introduced — built on top of a custom phase-aware data structure derived from a 750,000+ line JSON file.


Read Writeup
View Poster