Kale-ab is a Research Engineer with experience working on challenging theoretical and applied problems in deep learning and reinforcement learning. He considers an excellent work ethic, deep technical knowledge, and his ability to think innovatively as his greatest strengths.
His research interests include Multi-Agent Reinforcement Learning (MARL) (namely scalable coordination or value decomposition) and Deep Learning (specifically Pruning and Generalization).
He is also passionate about using technology to help the African continent. To this end, he has worked on projects that aim to have a high impact in Africa and has also worked on increasing diversity and representation in machine learning by serving on the steering committee of the Deep Learning Indaba.
For more information, you can view my resumé .
MSc in Computer Science, Focusing on Deep Learning (Distinction), 2018 - March, 2021
University of Witwatersrand
Honours in Computer Science (Distinction), 2016
University of Pretoria
BSc. in Computer Science, 2013 - 2015
University of Pretoria
Breakthrough advances in reinforcement learning (RL) research have led to a surge in the development and application of RL. To support the field and its rapid growth, several frameworks have emerged that aim to help the community more easily build effective and scalable agents. However, very few of these frameworks exclusively support multi-agent RL (MARL), an increasingly active field in itself, concerned with decentralised decision-making problems. In this work, we attempt to fill this gap by presenting Mava, a research framework specifically designed for building scalable MARL systems. Mava provides useful components, abstractions, utilities and tools for MARL and allows for simple scaling for multi-process system training and execution, while providing a high level of flexibility and composability. Mava is built on top of DeepMind’s Acme, and therefore integrates with, and greatly benefits from, a wide range of already existing single-agent RL components made available in Acme. Several MARL baseline systems have already been implemented in Mava. These implementations serve as examples showcasing Mava’s reusable features, such as interchangeable system architectures, communication and mixing modules. Furthermore, these implementations allow existing MARL algorithms to be easily reproduced and extended. We provide experimental results for these implementations on a wide range of multi-agent environments and highlight the benefits of distributed system training.
You can see a full list of my open source projects on github - https://github.com/KaleabTessera .