Bryan Oliveira
Hi! I'm Bryan, a 27-year-old AI enthusiast based in Goiânia, Brazil. I hold a Bachelor's degree in Computer Science from Federal University of Goiás, where I'm currently a Master's student focusing on Artificial Intelligence (AI) and Reinforcement Learning (RL). Elegant code writer, eclectic listener, casual photographer, voracious meta-learner, and sporadic gamer. I am perpetually fascinated by nature's wonders, especially the intricacies of the human mind.
I'm currently working at the Center of Excellence in Artificial Intelligence (CEIA) and the Advanced Knowledge Center for Immersive Technologies (AKCIT), leading AI research projects for real-time optimization, and Reinforcement Learning from Human Feedback (RLHF) to enhance large language models. My past experiences include leveraging RL for multichannel advertisement optimization and debt collection, and machine learning engineering for developing and deploying products for tourism recommendation, sales forecasting, and energy consumption optimization.
I'm deeply interested in artificial intelligence, neuroscience, high-performance computing, evolution, quantum mechanics, and their interconnections. I also like a broader philosophical approach to AI research, as I believe it can be used to explain numerous human behaviors and thought processes.
Sliding Puzzles Gym: A Scalable Benchmark for State Representation in Visual Reinforcement Learning
Workshop Publication · October 2024SPGym extends the 15-tile puzzle to evaluate RL algorithms by scaling representation learning while keeping latent dynamics and algorithmic problem fixed, revealing opportunities for advancing representation learning for decision-making research. Read more
Benchmarking Open-Source LLMs as Model Evaluators
Research Paper · October 2024We evaluate open-source LLMs against proprietary models using benchmarks for instruction adherence and positional bias, finding that open models are closing the performance gap with GPT-4, though GPT-4 still leads in overall consistency and fairness. Read more
Data-Driven Debt Pricing: A Systematic Literature Review
Research Paper · February 2023This review explores the potential of machine learning in debt pricing, with a focus on reinforcement learning. It concludes that more research is needed and highlights issues with reproducibility and comparability of results. Read more
PulseRL: Enabling Offline Reinforcement Learning for Digital Marketing Systems via Conservative Q-Learning
Workshop Publication · October 2021PulseRL is an offline reinforcement learning system for optimizing communication channels in Digital Marketing Systems (DMS) using Conservative Q-Learning (CQL). It learns from historical data, avoiding costly interactions, and reduces bias from out-of-distribution actions. PulseRL outperformed RL baselines in real-world DMS experiments, proving its effectiveness at scale. Read more
Multiagent Soccer Environment for Python
Reinforcement Learning Environment · September 2021A pre-compiled Soccer-Twos environment with multi-agent Gym-compatible wrappers and a human-friendly visualizer. Built on top of Unity ML Agents to be used as final assignment for the Reinforcement Learning Minicourse at CEIA / Deep Learning Brazil. Read more
Cellular Automata Framework
Project · March 2021A Cellular Automata program built with C++, OpenGL, CUDA and OpenMP. The main objective of this project is to allow scaling up to a reasonably large number of cells while maintaining the code legibility and allowing for further customisations. Read more
Intrinsic motivation for robotic manipulation learning with sparse rewards
Undergraduate Thesis · December 2019Intrinsic motivation for robotic manipulation learning with sparse rewards - Study of the impact of curiosity and intrinsic motivation as an exploration strategy for deep reinforcement learning agents on sparse-reward robotic manipulator environments. Read more
Bone Age Regression
Deep Learning · November 2019This is my code for the I2A2 Bone Age Regression competition. I learned a lot by building this pipeline from scratch and experimenting with different model architectures and optimizers. This was my first end-to-end image regression model, and it was very nice seeing my theoretical knowledge work in practice. Read more
Quack
Game · January 2019Quack is a Unity3D game made for the Global Game Jam 2019 themed "What home means to you?". The game consists of a happy chicken that wants to build a new home for its children. You have to collect sticks and group them on top of the main tree to make a lovely nest. This game was developed within 12 hours. Read more
3D Rendering & Force Simulator
Rendering · December 20183D Force simulator using only Processing's point() and line functions. Uses Digital Differential Analyzer (DDA) to render lines between two points, Scan Line to render polygons, normal calculation to determine faces to render in 3D space and Newtonian physics. Written in Java. Read more
IEEE VSSS Team
Robotics · October 2018A stack consisting of image processing, computer vision, team coordination, navigation, control and communication software to compete in the 2018's Latin-American Robotics Competition for the Pequi Mecânico UFG - INF's team. Read more
Die Zombit
Game · June 2015Reviving the classics of the 80's and 90's, Die Zombit is a retrowave top-down shooting game that has a striking soundtrack and an addictive gameplay which guarantee many hours of fun. Read more
