Hello!

I'm Tinuade (Tee-nu-ah-day), an AI research engineer interested in how language models learn, reason, fail, and change under different data, optimization, architecture, and objective choices.

I spent about three years building and deploying software across edtech, fintech, and developer tools, then completed an MSc in Artificial Intelligence at Heriot-Watt University.

Since then, I've become increasingly focused on AI research. I've participated in research fellowships including ARENA and APART, and continued developing as a researcher through Algoverse, where I mentor research projects and contribute to new research papers. My recent work spans logical reasoning in LLMs, uncertainty-aware evaluation, and steganographic behavior in language models.

My current research interests explore how model behavior and internal representations emerge from training choices. This includes questions around data quality, optimization theory, mechanistic interpretability, reasoning, and the science of deep learning.

I write about what I'm learning and my research process through "weekend-squashed" experiments that help me build intuition quickly. I also maintain a living research ideas document, where I collect questions I think about investigating.

Outside of research, I enjoy reading, running, and watching murder mystery and crime series. I'm fascinated by intelligence organizations and how they operate, though I do not support or endorse many of their actions. You can learn more about me through my scrapbook.

Feel free to say hi, send feedback on my work, suggest research ideas or tiny experiments, or reach out about collaboration.