Michael Wieck-Sosa

I am a fourth-year PhD Student in the Department of Statistics & Data Science at CMU. I am very lucky to be advised by Aaditya Ramdas.

Here is my CV. You can reach me at mwiecksosa AT cmu DOT edu.

My research interests are at the intersection of:

• Time series (high-dimensional, nonstationary, nonlinear, physical dependence, mixingales)
• Causality (structural causal models, invariant prediction, causal discovery, conditional independence testing)
• Machine learning (neural networks, transformers, foundation models, large language models)

Research

Under Revision

• Conditional independence testing with a single realization of a multivariate nonstationary nonlinear time series (with Michel F. C. Haddad and Aaditya Ramdas). arXiv.

Working Papers (Available Upon Request)

• Estimating dynamic models by matching random features (with Cosma Shalizi).

Works in Progress

• Transformers meet invariant prediction (with Aaditya Ramdas).
• Invariant prediction for time-varying sequences (with Aaditya Ramdas).
• Signature methods for large language models (with Leif Weatherby, Tyler Shoemaker, and Cosma Shalizi).
• Goodness-of-fit testing and confidence sets through random features (with Cosma Shalizi).
• Dynamic variable importance for time series foundation models (with Tom Zhang and Chad Schafer).

I have a few projects that involve pre-training transformers on synthetic time series data generated from neural structural causal models.

If you’re a CMU student (MS or advanced BS) and interested, send me an email.

News

• May 2026: I’ll be giving a 20-minute talk at the International Workshop in Sequential Methodologies at American University (June 1-4).
• April 2026: I was awarded the DeGroot-Goel Fellowship for 2026 by the Statistics faculty.
• April 2026: I successfully proposed my thesis, which is about prediction with transformers, estimation of time series models, and inference for dependence relationships.