The Acadev committee aims to make data science accessible by equipping students with practical skills and bridging the gap between classroom learning and real-world applications.
(01) — Our work
We lead and teach the DeCal course, “Introduction to Real World Data Science,” helping students of all experience levels build practical data science skills beyond the classroom.
Through a project-based curriculum, mentorship from our committee members, and an end-to-end team project using a real-world dataset, students gain the hands-on experience needed to tackle real data science and machine learning problems.
Our mission is to bridge the gap between classroom learning and industry practice, empowering students to build confidence and thrive in the rapidly evolving world of data science.
(02) — DeCal portfolio
Here are a selection of student projects developed with mentorship from Acadev instructors throughout the DeCal.
Built a computer vision system with MobileNetV3 and XceptionNet to distinguish real face images from AI-generated deepfakes. The project combined visual and frequency analysis with data augmentation, dropout, and early stopping to create a more robust detection pipeline.
Built an NLP-powered search tool that uses legal-text embeddings and similarity ranking to surface relevant court cases beyond simple keyword matches, with an experimental occlusion method for explaining which passages drive each result.
Analyzed historical player statistics to identify what separates NBA MVP candidates, then used Random Forest and Ridge regression models with feature reduction to predict and rank the leading contenders for the 2025 award.
Developed an interactive map of food desert risk scores across the United States, using machine learning and LLM-based analysis to help users explore geographic disparities in food access.
Compared pre-monsoon and early-monsoon transaction data to study shifts in spending behavior, then evaluated Random Forest and Logistic Regression models for predicting high-value purchases.
Explored how temperature, luminosity, and radius vary across different types of stars, then built a K-Nearest Neighbors model to classify stars from their physical characteristics.
Analyzed heart-failure clinical records to explore relationships between health indicators and mortality, combining visual analysis and chi-square testing with a K-Nearest Neighbors classification model.
(03) — Portfolio
Alongside our teaching responsibilities, our committee also works on a client project during the semester. Here are some of the projects that have helped our members gain more project experience.
Automated GPU trace analysis system for LLMs, enabling rapid detection of performance regressions and kernel fusion opportunities across software builds.
Interpretable ML framework predicting behavioral health treatment outcomes and revealing drivers of successful exits.
Interested?
We recruit in the first two weeks of Fall and Spring semester. Check the Join page for dates, timelines, and how to apply.
