Two sides of my work: data science projects with their results, and apps I have built.
Seven projects from my MSc in Data Science, each with the question, the method, the results and what I would do next.
Machine learning
Supervised, unsupervised and NLP models on real datasets
Machine learning
71%Buyer recall after tuning (from 35%)
Online Shopper Purchase-Intent Prediction
Predicts which e-commerce sessions end in a purchase. Tuning for the minority class lifted logistic-regression buyer recall from 35% to 71% on 12,330 sessions.
Classifies 50,000 movie reviews as positive or negative with TF-IDF features and linear models, reaching 90.1% accuracy and 0.966 ROC-AUC on 10,000 held-out reviews.
Unsupervised segmentation of 2,111 people by eating habits, activity and physical profile. K-Means outperformed hierarchical clustering on all three validity metrics.
A collaborative-filtering recommender built with Spark MLlib on 200,000 Steam purchase and playtime records, tuned across eight ALS configurations and tracked with MLflow.
Spark SQL analysis of more than 520,000 registered clinical trials on Databricks: study-type mix, most-researched conditions, average trial duration and the rise of diabetes research.
Foreign keys over five imported CSV tables, then ten T-SQL answers to business questions: subqueries, joins, aggregation, TOP-N ranking, VAT arithmetic and a discount procedure.
My Fildata dissertation and internship work is covered by confidentiality agreements, so it is described on the Experience page rather than shown here.
Apps I have worked on, across commerce, mobility, finance, health, social and more. Each one has a short overview and a Demo button that opens that app’s own walkthrough video on Google Drive.
