Hi, I'm Ibra Niang. I work on multimodal NLP: vision language models and the systems that train them. I am finishing a combined BS/MS at the University of Memphis in 2026, advised by Dr. Xiaolei Huang.

Most of what I do sits close to the metal. I study post-training and distillation, run scaling-law sweeps, and build training and inference stacks from scratch in JAX, CUDA, and C. I like questions whose answer is a number you have to earn, so I tend to write the trainer, run the sweep, and then publish what actually happened, including the runs that did not work.

I live in Memphis, Tennessee.

My code is on GitHub, my background is on LinkedIn, and shorter notes go on Twitter. You can reach me at [fname][lname]998[at]gmail.com.