Sheila Acar
Turning complex data into clear, actionable insights. MS in Data Science from Northwestern, specializing in AI, machine learning, and data visualization.
The Person Behind
the Data
Hi, I'm Sheila Acar — a data analyst based in Chicago who believes the best analysis is one that drives real decisions.
I currently work as a Data Analyst at the American Society for Surgery of the Hand, where I conduct ad hoc analysis, built the organization's first live dashboard for real-time operational visibility, and design surveys to improve member experience and retention.
I hold a Master's in Data Science with an AI Specialization from Northwestern University (GPA 3.95), where I built a deep foundation in machine learning, NLP, statistical modeling, and data visualization.
Outside of data, I'm passionate about community — I served as a Student Liaison at Northwestern's Student Leadership Council, connecting students and shaping the graduate program experience.
Selected Work
Board Game Recommendation Chatbot
Built and compared two end-to-end chatbot systems for recommending board games from a cleaned 8,000+ game dataset sourced from BoardGameGeek. The first approach uses LlamaIndex + OpenAI GPT for semantic vector search; the second deploys Llama 2 locally with TF-IDF retrieval and cosine similarity for context-aware, multi-recommendation responses. Evaluated both models across five query types — Llama 2 consistently delivered more relevant, detailed answers.
BoardGameGeek Data Cleaning & NLP
Cleaned and enriched a 100,000+ game dataset from BoardGameGeek — removing outliers, handling missing data, and scraping game summaries from BGG, Wikipedia, and Google. Applied NLP preprocessing (tokenization, lemmatization, stopword removal) to prepare the corpus for downstream modeling.
Board Game EDA & Market Analysis
Conducted exploratory data analysis on 8,000+ board games from BoardGameGeek to uncover what drives high ratings and widespread ownership. Identified a positive relationship between complexity and rating averages, found that top-rated games skew toward niche wargames with hexagon grids and dice rolling, and determined that the most-owned games are newer strategy and family titles involving drafting and cards. Findings informed a hypothetical board game design framework based on target audience goals.
Predicting ADHD & Sex from Brain Imaging Data
Capstone project developed for the 2025 WiDS Datathon. Built and compared multiple ML models — Random Forest, SVM, Logistic Regression, and Graph Neural Networks — to classify ADHD diagnosis and biological sex using fMRI connectome matrices and socio-demographic, emotional, and parenting data from 1,200+ child and adolescent subjects (Healthy Brain Network / Child Mind Institute). Applied SMOTE for class imbalance, NetworkX for graph feature extraction, and both RFE and SHAP-based feature selection. Random Forest with SHAP features achieved 90% ADHD accuracy and 82.71% sex classification accuracy, outperforming all GNN architectures.
CNNs for Colorectal Cancer Histopathology
Designed and evaluated 18 convolutional neural network architectures to classify 8 tissue types in colorectal cancer histology images using the TensorFlow colorectal_histology dataset (5,000 images). Systematically explored dense layer sizes, hidden layer depth, dropout rates (0.2–0.4), L2 regularization, batch normalization, augmentation, and optimizer choice (Adam, RMSprop, SGD). The best model — 3 conv/max-pooling layers (32, 64, 128 filters), 128 dense units, and 0.2 dropout — achieved 77.8% test accuracy with closely matched training and validation curves, indicating strong generalization.
Background & Experience
My journey into data science didn't start in front of a screen. It started in a lab. Before making my career transition in 2025, I worked as a Research Technologist performing histological processing on animal tissues for cancer research. Working as a histology technician in a research setting gave me thoroughness, patience, and an eye for detail that translates directly into how I approach data.
At some point, I realized that same mindset translated beautifully into data science and I decided to make the leap. I completed my Master's degree in 2025, and having that scientific foundation made the journey richer. I bring something a little different to the table: a background that blends hands-on research experience with modern data science skills, and a genuine appreciation for how rigorous methodology leads to results you can actually trust.
Download My Resume
Full work history, education, and skills in one PDF.
Download PDF Last updated: April 2025Let's Work Together
I'm open to new roles, freelance projects, and interesting data collaborations. Based in Chicago and always happy to talk data, analytics, or AI — reach out anytime.