Portfolio

(2021) Table Tennis Equipment Recommender: Distance-based recommender engine for rubbers. Python model built using PCA and KNN, deployed into Flask API. Model API consumed with a React application.
Demo: Link
Post: Link
Skills: Sklearn, Unsupervised Learning, Flask, React.js, React Boostrap, Axios React  

(2021) Text Classification on Table Tennis Rubbers: Text Classification with Scikit-Learn. Extracted data with Selenium and hand-labeled examples to build a classifier. AUC: 0.957 and Accuracy: 0.87.
Post: Link
Skills: Sklearn Pipelines, Text Classification, Text Features, NLTK, Spacy  

(2019) Abstractive Text Sumarization: Abstractive text summarization. Used an Encoder-Decoder arquitecture with attention and pre-trained embeddings (GloVe). Beam search used for inference. Rouge scores: Rouge-1, 23.27; Rouge-2, 15.30; Rougue-L: 24.47.
Notebook: Link
Skills: Word Embeddinigs, Tensorflow 2, Sequence Modeling  

(2019) Extractive Text Sumarization App: Extractive text summarization app using TextRank with Gensim. Deployed to heroku with Flask.
Demo: Link
Code: Link
Skills: NLP, Flask.  

(2018) DonorChoose.org Application Screenig: A classification model using text features to predict whether or not a DonorsChoose.org project proposal submitted wiil be approved.
Notebook: Link
Skills: Feature Engineering, Text manipulation, LightGBM  

(2018) Predicting Boston Housing Prices: A model for price prediction using tree-based models. Bayesian Optimization used for Hyperparameter tunning on XGBoost.
Notebook: Link
Skills: Data manipulation, Random Forest, XGBoost