Hector Motsepe bio photo

Email

My CV

Twitter

LinkedIn

Github

About Me

Here is Hector Motsepe.

I am a Machine Learning Engineer at Lexis Nexis AI. I hold a Master Degree (Deep Learning and Machine Learning) from University of Pretoria, advised by Prof. Aurona Gerber. I have been awarded Huawei Top Achiever Scholarship

If you are interested in any aspect of me, I would love to chat and collaborate, please email me at - hectormotsepe[at]gmail[dot]com

Academic Background

[Highlight] I am actively looking for PhD position to start in 2024 Fall. Contact me if you have any leads!

  • Sep 2021 - June 2023: Unveristy of Pretoria (MS, CS)
  • Jan 2020 - Dec 2020: Univeristy of Johannesburg (Bcom, IS)
  • Jan 2017 - Dec 2019: Unveristy fo Johannesburg (Bcom, IS)

Research Interests

  • Deep Learning (NLP/CV)
  • Reinforcement Learning
  • Backpropagation (Straight Through Estimator)
  • Applied Machine Learning
  • Proposals Coming Soon !!! [My first research proposal], [My second research proposal]

My research focuses on developing a modular cognitive architecture for autonomous agents that can continuously update its world model through experience and effectively integrate feedback across different modules. Key objectives include enabling bi-directional communication between the world model and low-level controllers, facilitating coordination among modules for task execution, and incorporating multi-modal inputs to enhance the environment representation within the world model. The overarching goal is to advance autonomous agents that can acquire new knowledge through experiential learning and adapt their behavior in unfamiliar situations by leveraging an evolving, comprehensive understanding of the world around them.

My second research interest lies in exploring techniques to enable effective backpropagation through the discrete sampling process in large language models. Specifically, I aim to advance the straight-through estimator (STE) or develop novel methods to propagate gradients during the multi-step text generation process. Overcoming this challenge could unlock significant improvements in training large language models, leading to better performance in natural language processing tasks such as text generation, summarization, and machine translation. Ultimately, my goal is to contribute to the development of more capable and versatile language models that can positively impact a wide range of applications and benefit society.

Continuous Learning

I have created a YouTube Channel to share my knowledge, and my understanding of deep learning concepts

News and Updates

  • **Nov 2023: **Joined Lexis Nexis AI as a Machine Learning Engineer, contributing to cutting-edge projects in AI technology.
  • **Aug 2021: Joined Integrove (client: Anglo American Group), South Africa, Johannesburg: Spearheaded the implementation of machine learning models, revolutionizing variance analysis calculations and drastically reducing code complexity. Led a successful migration project to transition from on-premise infrastructure to the cloud, resulting in a significant cost reduction.
  • **Jan 2021: Worked as a Data Engineer at PBT Group Pty Ltd. (client: Old Mutual Insurer), South Africa, Johannesburg: Implemented distributed processing pipelines using Spark, achieving a remarkable 67% improvement in data processing time. Designed and optimized data engineering pipelines, streamlining data migration processes and eliminating manual uploads.
  • 2019-2020: Served as a Software Engineer (Machine Learning) at Alpexes Pty Ltd., focusing on designing high-performance backend systems and integrating machine learning algorithms.