Excelerate × Rochester Institute of Technology
AI-Powered Data Analysis Internship
Star Performer work across data understanding, cleaning, visual analysis, predictive modeling, and recommendations for RIT student engagement.
360° Evaluation
Three perspectives · averaged- 84%
- Self
- 100%
- Peer
- 100%
- Managerial
- 95%
- Overall average
8,558
Student records analyzed
8,024
Records after cleaning
86.1%
Model accuracy
95%
Overall internship score
Problem & Context
Rochester Institute of Technology (RIT) and Excelerate wanted to understand what drives students to participate in opportunities after signing up. With thousands of records and many potential variables, the challenge was to move beyond guesswork and identify the real levers of engagement.
My task was to clean the dataset, explore patterns, build a predictive model, and present findings that could inform program design and outreach strategy.
Dataset & Methods
- Dataset: 8,558 RIT student engagement records with demographic, behavioral, and opportunity-type fields.
- Cleaning: Removed 534 records with missing learner or opportunity identifiers, removed four non-analytical or misleading columns, and engineered 14 fields. The model-ready dataset had 8,024 records across 26 columns with zero missing values.
- Exploration: Cohort segmentation and comparative analysis across opportunity categories, signup timing, and demographic groups.
- Modeling: Built and evaluated a Random Forest classifier to predict participation outcomes and rank feature importance.
Analysis workflow & evidence
Follow the work from the first data-quality review through cleaning, charts, modeling, and the final recommendation.
01 · Understand
Dataset structure and data quality
The initial review documented the dataset, variables, missing entries, and date-format issues before analysis began.
Open Week 1 data-understanding report02 · Clean and explore
Cleaning, feature engineering and visual insights
The Week 2 report records the transition from 8,558 raw rows to 8,024 records across 26 columns, with zero missing values, plus the exploratory analysis and charts.
Open Week 2 visual-insights report03 · Model
Prediction and model evaluation
The Random Forest report includes evaluation metrics, feature-importance visuals, a confusion matrix, and a predicted-participation heatmap.
Open Week 3 prediction report04 · Present
Final program insights presentation
The final presentation summarizes the analysis and recommends a real-time participation-probability dashboard for program managers; it is a proposed next step, not a claim that a live dashboard was deployed.
Open final presentation
Key Findings
Opportunity Category is the dominant predictor
Random Forest feature importance showed that the type of opportunity — not region, age, or gender — most strongly predicted whether a learner would participate.
Events outperform internships by a wide margin
Events held a 96% active-participation rate, while internships showed a 65% rejection rate, suggesting very different learner intent and friction points.
Engagement timing predicts retention
Learners who applied 1–3 months after signup outperformed same-day applicants by 20 percentage points, indicating that rushed onboarding may lower commitment.
Recognition
This internship earned Star Performer award for the Excelerate × RIT AI-Powered Data Analysis internship, with a 95% overall score, including 84% self-evaluation, 100% peer evaluation, and 100% managerial evaluation.
Let's Talk
Need data-driven operations support?
I turn messy data into clear decisions — and clear decisions into systems that save time.