(pronounced Finegold)
I'm currently working on a series of personal projects to deepen my understanding of Data Science and Statistics, using both Python and R. My focus has been on building practical tools like a next-word prediction model and visualizing natural language data.
Before pivoting into data science, I spent over two decades in web and mobile startups, building B2B and consumer-facing products. My work included:
- Early 2000s: Secure document collaboration platforms and e-learning systems for continuing education credits
(ASP, .NET, C#, JavaScript, SQL) - 2010s: Two mobile social networks — one for short-form video (iPhone 3 and early Android), another in the dating space
(Objective-C, Java, PHP, SQL)
Now, I'm bringing that product and engineering background into the world of data.
A Shiny web application that predicts the next word in a sequence of English text using Kneser-Ney-smoothed n-gram language models (bigrams to fivegrams). The app includes:
-
An interactive UI with real-time predictions
-
A backend powered by preprocessed corpora and optimized SQLite queries
-
Optional Markov chain visualizations of text structure
-
Perplexity metrics for evaluation
🖥️ Live App: Next Word Predictor on Shiny
🧾 Project Presentation: Slides
Technologies: R, Shiny, SQLite, tidyverse, textclean, textstem, Kneser-Ney smoothing
What I Learned:
-
Higher-order n-grams (four- and five-grams) are rarely useful in practice, supporting Markov’s insight that language is mostly driven by short-term dependencies.
-
Efficient text preprocessing and query optimization were essential to making the app run on limited hardware (8GB RAM, 2017 MacBook Air), which involved aggressive normalization, indexed lookups, and careful memory use.
Explore and compare disc golf courses across the U.S. using interactive maps and visualizations.
-
Built with R Shiny, Leaflet, and ggplot2 to create an intuitive app for browsing, filtering, and analyzing over 7,000 disc golf courses.
-
Users can:
-
Filter courses by state
-
View location details and links via popups
-
Jump to a random course with a smooth map animation
-
Compare course ratings to state or national distributions using dynamic histograms
-
-
Integrated external datasets and designed custom icons for a polished user experience.
🔗 Live App
📊 Dataset on Kaggle
Built a supervised machine learning model to predict the success of SpaceX Falcon 9 launches.
-
Explored and visualized SpaceX launch data using pandas, matplotlib, seaborn, and plotly.
-
Engineered features from categorical launch attributes including launch site, payload mass, and booster version.
-
Trained and evaluated several classification models (Logistic Regression, Random Forest, SVM, KNN) to determine the likelihood of a successful landing.
-
Tuned hyperparameters and assessed performance using metrics like accuracy, F1 score, and confusion matrix.