I'm Alejandra Berbesi

I love learning new technologies. I'm interested in education projects & tech for good causes 🦕

Add me in linkedin

View My GitHub Profile

Welcome to my Portfolio

These are some of my projects:

Scratch Game

Designed for kids (in Spanish) using Scratch (a block-based visual programming language). The goal is to help a diver clean the ocean, picking up plastic garbage.

:smiley_cat: Link to Game

'Tu Libro A Ciegas' model

NLP/transfer learning to classify literary categories of books. Technologies involved: Python (Transformers, Pytorch, Pandas, Scikit-learn, NumPy), Google BigQuery.

:books: Github_Repo

Spanish Language model

122000+ wikipedia spanish articles to create a language model (next word prediction) that could be later used for transfer learning. Technologies involved: Python (Fastai2).

:book: Github_Repo

Multi-label apparel images prediction

Identifying type-color combination for 5 types and 6 colors of clothing’ images using a convolutional neural network (CNN). Technologies involved: Python (Fastai).

:kimono: Github_Repo

'The Simpsons' image classification

Identifying 42 characters of ‘The Simpsons’ using a convolutional neural network (CNN) backbone and a fully connected head with a single hidden layer as a classifier. The output is the predicted probability for each of the categories. Technologies involved: Python (Fastai).

:tv: Github_Repo

Emotions in tweets

Script that classifies tweets in 13 different emotions. ‘Master’ branch: a model using a random forest as a classifier. ‘Neural network’ branch: the model is built with a Long Short Term Memory (LSTM) neural network. Technologies involved: Python (Pandas, Numpy, Scikit-learn, Nltk, tensorflow-keras, Matplotlib).

:speech_balloon: Github_Repo

Sentiment analysis (movie reviews)

Script to evaluate reviews about movies, classifying them with positive or negative polarities. ‘Master’ branch: a model built with Support Vector Machine. ‘Naive_bayes’ branch: a model with a class that returns the result of a voting process between Naive Bayes, Multinomial Naive Bayes and Bernoulli Naive Bayes classifiers. Technologies involved: Python (Numpy, Scikit-learn, Nltk).

:movie_camera: Github_Repo