Jaeyoung Kim

I’m Jaeyoung Kim, a researcher and CTO working at the intersection of machine learning and human-centered AI.

My recent work focuses on evaluating and interpreting how vision-language and multimodal models actually behave — auditing whether their reported capabilities reflect genuine reasoning or memorized priors, and whether the mechanisms they claim to use are actually active. I’m drawn to questions that sit underneath the benchmark numbers: not just how well a model scores, but why, and whether we should trust it.

I bring a background in visualization and human-computer interaction to this, which shapes how I work — I try to turn opaque model behavior into diagnostics that are both technically rigorous and interpretable to people.

Currently I’m the CTO of MADI Inc., a healthcare-AI company, where I lead applied ML development while keeping an active research footing. Before that, I completed my Ph.D. in Computer Science at Seoul National University.

Feel free to reach out — the best way is via email or LinkedIn.