I am John Olafenwa, a machine learning engineer focused on Large Language Models and Reinforcement Learning.

I work on post-training of reasoning models at Microsoft, and in my personal time, I like to study information theory, read papers on RL and make YouTube videos about large language models.

About

From self-taught developer to ML engineer

I started my career as a self-taught developer. In 2017 I taught myself machine learning and deep learning, spent some time obsessing over optimizing the efficiency of CNNs, and founded a startup with my brother focused on building and deploying computer vision models on the edge.

I joined Microsoft Mixed Reality in 2020 to work on object and human body tracking from vision and audio signals. Many years later, I have shipped models running within Microsoft Teams to over 300 million people.

I have been a strong believer in AGI for many years, and I spend most of my time today either post-training reasoning models at Microsoft, or studying information theory, reading papers on RL, and making YouTube videos about large language models.

I love tackling new challenges, and my philosophy is “anyone can learn anything.”

My core interests lie in alignment science and understanding the nature of learning systems. I am a lifelong learner, and it makes me curious about how learning happens in both humans and agents. I aim to figure out how to leverage what we know about teaching humans to efficiently train and align AI systems.

Presentations

YouTube videos

Long-form explanations of the maths and code behind modern training methods.

Projects

Things I have built and released

Newsletter

Substack

Essays on AI, reasoning and the changing world of work.

Articles

Medium

Essays, paper reviews and tutorials.

Contact

You can reach me by email or on LinkedIn.