Marcos Pereira, Machine Learning Developer in Lisbon, Portugal
Marcos Pereira

Machine Learning Developer in Lisbon, Portugal

Member since July 4, 2017
Marcos is an experienced software developer and machine learning researcher. Having worked with all kinds of companies—from VC-funded startups to established corporations—he believes code is a means to an end and also sees it as an art when it comes to making it easy for others to build upon it. As a member of the machine learning collective research lab, he collaborates with fellow researchers whose backgrounds include places such as Google Brain and Uber AI Labs.
Marcos is now available for hire

Portfolio

Experience

Location

Lisbon, Portugal

Availability

Part-time

Preferred Environment

Windows, Visual Studio Code, GitHub, Unity, Jupyter, Google Colab

The most amazing...

...project I've built is a neural network inspired by cellular automata that learns to solve reasoning tasks.

Employment

  • Researcher

    2020 - PRESENT
    Machine Learning Collective
    • Built a deep learning model based on neural cellular automata that learned to solve reasoning tasks.
    • Provided feedback to other members on research group meetings.
    • Participated in discussions on future research projects.
    Technologies: Jupyter, Python, TensorFlow, Keras
  • Back-end and Data Engineer

    2019 - 2020
    Athlane
    • Built the API that acted as the company's backbone for managing streamer and sponsorship data.
    • Built various tools for data gathering and analysis.
    • Led back-end development team, organizing server architecture and distributing the workload.
    Technologies: JavaScript, Node.js, Express.js, PostgreSQL
  • Back-end Developer

    2019 - 2020
    Youblicity
    • Built a Node.js API to support front-end functionality such as user management, sponsor registration, and payment tracking.
    • Built various tools to handle Twitch data extraction and processing.
    • Used BigQuery to extract useful analytics from several terabytes of streamer data.
    • Built computer vision tools to reward sponsored Twitch streams for partnership compliance.
    Technologies: Node.js, BigQuery, Data Science
  • Lead Engineer

    2018 - 2019
    Swish Labs
    • Delivered a popular browser extension to thousands of active users.
    • Led the tech overhaul of an eastern market leader clothing label.
    • Developed a secure and concurrent share trading market from scratch.
    • Developed self-documenting REST and GraphQL APIs backed by Node.js.
    Technologies: MongoDB, PostgreSQL, Vue.js, Node.js
  • Web Developer

    2018 - 2018
    Kogneta
    • Created internal tools for improving reach and customer acquisition.
    • Improved customer onboarding efficiency.
    • Leveraged Google Cloud solutions for data processing.
    • Developed efficient serverless tools for data storage and analysis.
    • Developed additional functionalities for Google Data Studio for internal use.
    Technologies: Google Cloud, Node.js, Serverless Framework
  • Tooling Engineer

    2017 - 2018
    Kitty's Co.
    • Contributed to internal tools for optimizing the work of the agency's creatives.
    • Developed Node.js scripts that interacted with the Facebook Marketing API.
    • Created serverless scripts that were run on AWS Lambda.
    • Wrote AWS Step Functions to manage and coordinate script execution.
    Technologies: AWS Step Functions, AWS Lambda, Facebook Marketing API, Node.js
  • Lead Developer

    2014 - 2017
    Okorana
    • Created a drag and drop interface in AngularJS that allows for placing and resizing of widgets.
    • Developed REST APIs used to power single-page apps using both SQL and NoSQL databases.
    • Set up build systems using Grunt and later on Gulp.js, increasing useful development time.
    • Led a team of developers, organizing contributions and communication between members.
    • Hired new team members to assist with the front-end design of web apps.
    Technologies: MongoDB, AngularJS, Node.js, JavaScript

Experience

  • Neural Cellular Automata for ARC

    A deep learning research project where a model was trained to solve tasks from François Chollet's Abstraction and Reasoning Corpus. The model was based on and inspired by previous work on neural cellular automata.

  • Dead Drive

    A multiplayer, open-world, survival game where players build a base on a moving truck while avoiding a constantly following horde. Built using Unity, this game is scheduled for release in Early Access on Steam in the second half of 2021.

  • Backpropagation From Scratch
    https://github.com/marcospgp/backpropagation-from-scratch

    A convolutional deep learning model built from scratch using Python and Numpy.
    The neural network features two hidden layers with sigmoid activations on all layers except the last, which applies softmax. The model achieved 85% accuracy on MNIST with no regularization or data preprocessing.

  • CharityTab

    CharityTab was a Chrome extension and web app that allowed users to customize their browser's new tab page while donating to a charity of their choice by viewing ads and shopping through sponsored bookmarks.

  • Twitter Infinite Scroll Disabler
    https://github.com/marcospgp/twitter-infinite-scroll-disabler

    A productivity tool that disabled Twitter's infinite scrolling feed and replaced it with a "Show More" button. This project was deprecated due to the high amount of maintenance required to keep up with regular updates to Twitter's web user interface.

  • Twitch Bot
    https://github.com/marcospgp/koalabot-custom-modules

    A set of custom modules written for a popular Twitch bot that allows users to duel and gamble the points they earn while watching their favorite streamers. Developed in JavaScript and deployed using Heroku.

  • Infinifty Web Apps
    https://github.com/marcospgp/infinifty-web-apps

    A link shortener and web status checker built in PHP. The link shortener was built to encompass two domains: the host domain that allows the user to minify links and a short domain that redirects the shortened addresses to the correct endpoint.

  • Wareztogether
    https://github.com/bfar97/wareztogether-chrome-extension

    A Chrome extension that synchronizes video players between users of a popular streaming website. Users join a room through the extension and have their video player automatically synchronize with the players of everyone else in the room going forward.

  • Flappy Goat
    https://github.com/marcospgp/flappygoat

    A Flappy Bird clone in JavaScript, built from scratch with a custom HTML5 game engine as a way to work on my game development skills.

  • The Big Deuce
    https://github.com/marcospgp/big-2

    The popular Big 2 card game developed in C with an HTML-based CGI (Common Gateway Interface) server. This project features a medium difficulty AI opponent which acts on behalf of each of the other three players.

  • Haskell Sokoban
    https://github.com/marcospgp/haskell-sokoban

    A 2D puzzle game where the player pushes crates around in a warehouse, trying to get them to the correct storage locations. This project was written in Haskell and uses the Gloss vector graphics library.

Skills

  • Languages

    TypeScript, SQL, C#, GraphQL, C, Java, Python, JavaScript, Haskell, PHP, Go, HTML
  • Frameworks

    Unity3D, Unity, AngularJS, Express.js
  • Libraries/APIs

    NumPy, Node.js, Keras, TensorFlow
  • Tools

    BigQuery, Jupyter, GitHub
  • Platforms

    Docker, AWS Lambda, Visual Studio Code
  • Storage

    MongoDB, MySQL, PostgreSQL
  • Other

    Chrome Extensions, Algorithms, Machine Learning, Google Colab, Game Development, Deep Learning
  • Paradigms

    Data Science

Education

  • Bachelor's degree in Software Engineering
    2015 - 2018
    University of Minho - Braga, Portugal

To view more profiles

Join Toptal
Share it with others