Mario Toledo

Full Stack Developer, Computer Scientist and Gaming Lover

About me

Hello, my name is Mario, and I work professionally with software development for more than 10 years. Currently, I'm focused on Web Development using mainly JavaScript and Python, and Game Development using Unity, but I also have professional experience with mobile apps and other more corporative software.

I'm also a Computer Scientist and a M.S Computer Engineering, with published papers on scientific events. My current studies involves Machine Learning, Big Data, Computer Vision and Microservices.

And, oh... I really like Pokémon! <3

My Squirtle t-shirt (drawn by Kyuu) My Squirtle t-shirt (drawn by Kyuu)

Personal Projects



A community/hub for players to connect and interact

Founded in 2010, N-Party is a community to players to connect and interact with other players by throught local and digital experience. By using NParty website, the users can lookup for groups of people to play video-game, join or manage championships, and even write news and articles about the video-game industry. The user has a gamefied profile, gaining experience, levels and badges by interacting with the group experience.

Browse Website


Podcast about japanese pop culture

Subarashow brings weekly podcasts dedicated japanese pop culture in a fun and informative way, using the best of young language to make the viewer feel inside a casual conversation between friends.

The project contains a CMS for podcasts and an API built with Meteor JS, while the front-end was build using Node and Express JS.

I'm also one of the hosters and editor of most of episodes.

Browse Website

Other Projects

Some other of my personal projects
Some other of my personal projects

Other Projects

Some other of my personal projects

Professional Experience

Kokku (nov/2022 - current)

Engineering Management for one of the biggest gaming co-development companies in Brazil.

Afterverse (jun/2021 - oct/2022)

Lead game development using Unity for PK XD, an online mobile game with 50kk MAU.

Wildlife Studios (ago/2020 - jun/2021)

Full-stack development using Javascript and Python, and game development using Unity.

Ortiz Gaming (jul/2019 - ago/2020)

Development of HTML5 games using Javascript with Phaser.JS.

2Mundos Inc. (nov/2016 - apr/2019)

Full-stack development using mainly Javascript for back-end and front-end, alongside with technologies and frameworks like Node, Angular, Meteor, Ionic, and MongoDB.

Eixo X (aug/2013 - nov/2016)

Software development over several technologies, from web projects to native apps, using C#, Java, Objective-C and Javascript, alongside with technologies like MSSql Server, MySql, Node and more.

Glambox Brasil (feb/2013 - aug/2013)

Web development for an e-commerce using .NET, MS Sql Server and more. Helped on several internal solutions, building content management systems, stock systems, subscription systems and more.

Digital Pages (feb/2011 - feb/2013)

Software development of native applications on iOS using Objective-C. Mostly of apps were digital readers for famous magazines and newspaper from Brazil, like Folha de S. Paulo, Caras, Valor Econômico and more.

Insolita Studios (oct/2010 - fev/2011)

Development of games using C# and Unity engine. Also helped to create a framework for 2D development on Unity.

IBM (nov/2008 - jul/2010)

Software Development of corporatative projects using ASP, COBOL and Oracle Database. Also helped on software requirements and requirements elicitation.

Academic Experience


Instituto de Pesquisas Tecnológicas do Estado de São Paulo (IPT-SP)
Master's Degree, Computer Engineering
Master Thesis
Centro Universitário Senac
Bachelor's Degree, Computer Science
Undergraduate Thesis

Published Works

ISBN: 9786525263274

Cold-Start Problem é um problema recorrente em Sistemas de Recomendação nas seguintes situações: quando um novo item é adicionado ao sistema e não possui nenhuma avaliação prévia; ou quando um usuário sem histórico de avaliação entra no sistema. Avaliando as diferentes situações em que o Cold-Start Problem se apresenta, é possível considerar o uso do histórico de navegação como alternativa para geração de recomendações. Levando em conta o formato sequencial dos dados, estudos sugerem o uso de Redes Neurais Recorrentes (RNN) por permitir maior entendimento da sequência de dados e seu contexto. Durante a revisão sistemática realizada neste trabalho, as arquiteturas de LSTM, GRU e híbridas aparecem com frequência entre as pesquisas relacionadas ao tema. Entretanto, os autores dos trabalhos revisados não comparam as arquiteturas entre si, o que é crucial para o entendimento das vantagens e desvantagens do uso de dados do histórico de navegação com RNN. Este estudo propõe a comparação das arquiteturas de LSTM, GRU e híbridas de RNN através da criação de protótipos utilizando a mesma base de entrada, avaliando suas performances através dos valores de Acurácia, Revocação, Precisão e F1-Score..

october 2020
ICAAI 2020: 2020 The 4th International Conference on Advances in Artificial Intelligence, London, United Kingdom, October 2020 - DOI:

This article shows the results of a performance analysis from LSTM, GRU and Hybrid Neural Network architectures in Recommendation Systems. To this end, prototypes of the networks were built to be trained using data from the user's browsing history of a streaming website in China. The results were evaluated using the metrics of Accuracy, Precision, Recall and F1-Score, thus identifying the advantages and disadvantages of each architecture in different approaches.

may 2018
15th CONTECSI - International Conference on Information Systems and Technology Management ISSN 2448-1041 - DOI:

This work presents a research’s result that had the objective of analyzing the data of car robbery in São Paulo in a certain set of days, in order to identify a pattern among the robberies events presents in the data sample. Data were obtained from the São Paulo State Government’s Transparency Portal, using K-Means and Apriori algorithms for classification and grouping.

october 2015
WVC 2015. XI Workshop de Visão Computacional. p. 89 - ISBN: 978–85–8023–032-1

This work presents the results of using an Android phone as a control unit for a Lego NXT Robot. The smartphone’s camera allows the usage of Computer Vision techniques to allow the robot control by a superior processing unit than the default NXT CPU.