Silvio Mori Neto
I am a software engineer passionate about cutting-edge technology. I am keen on expanding the horizons of Artificial Intelligence so as to enhance customers' experience. I eagerly look forward to developing innovative solutions for big real-life problems, from small smart devices to complex integrated systems.
Portfolio
On this page, I expose some projects I have been developing to broaden my technical skills.
If you have some questions or comments, I am thrilled to expand the subject, share experience, and learn from you.
Machine Learning
Recommendation System
The problem this project propoes to solve is finding the most appropriate offer for each one of the customers, which means finding the offer that is more likely to lead the customer to buy Starbucks products.
Sentiment Analysis
This project uses techniques of Natural Language Processing as well as Recurrent Neural Networks to perform sentiment analysis on movie reviews. It also provides a publicly accessible API deployed on Amazon AWS.
Plagiarism Detection
By using NLP techniques, this project examines a text file and performs binary classification labeling that file as either plagiarism or not, depending on how similar that text file is to a provided source text.
Fraud Detection
This project analyzes a credit card fraud detection dataset so as to build a binary classification model that can identify transactions as either fraudulent or valid, based on provided, historical data.
Deep Learning
Flower Species Classifier
The challenge in this project is building a deep learning model, by using Transfer Learning, that identifies 102 species of flowers from images. The final application receives an image provided by the user and predicts the likelihood of being one of those known species.
Dog Breed Classifier
In this project, a Convolutional Neural Network (CNN) is built from scratch. The final model uses Transfer Learning. Given an image of a dog, the model identifies an estimate of the canine’s breed. If supplied an image of a human, it identifies the most resembling dog breed.
Human Face Generation
This project defines and trains, a Deep Convolutional Generative Adversarial Network (DCGAN) on a dataset of human faces. The model draws, by itself, brand new images of human faces that look so realistic that another model cannot tell real from generated faces.
TV Script Generation
This project generates its own Seinfeld TV scripts using Recurrent Neural Network (RNN/LSTM). The model is trained on the Seinfeld dataset of scripts from nine seasons, so that it is able to write a brand new TV script, based on patterns it recognizes in this training data.
Deep Reinforcement Learning
Collaboration-Competition
The interesting point in this project is that two completely separated models are trained, one against the other, to play tennis. They must learn the expected actions from their adversary in order to keep the game rolling. This is an implementation of the Multi-Agent DDPG algorithm.
Continuous Control
This project train one single brain to control up to 20 double-jointed arms simultaneously. The arms have to keep their hands in contact with a moving target (the green balls) for as many time steps as possible. This project implements the DDPG: Deep Deterministic Policy Gradient algorithm.
Navigation
The goal in this project is to create and train an agent to navigate and collect yellow bananas in a large square world, avoiding the blue ones. This project implements some of the cutting-edge deep reinforcement learning algorithms, such as Deep Q-Learning and Prioritized Experience Replay.
AWS DeepRacer League
In this challenge, the main goal is define a loss function and train an autonomous car by using Reinforcement Learning so as to race in virtual tracks. I was one of the best-ranked competitors out of thousands, thus I was offered a full scholarship to the Udacity Machine Learning Nanodegree Program.
Side Projects
Covid-19 Data Lake
In this project I built the cloud architecture for an ETL pipeline so as to collect information about the COVID-19 from distinct sources. This architecture was built by using Airflow, Spark, Docker container, AWS EMR, AWS ECS, AWS CloudWatch, and AWS S3.