Tower was pleased to welcome Andrew Wilson to our New York office last month for a seminar on advances in foundation models and data generation for time-series forecasting. A professor at NYU’s Courant Institute of Mathematical Sciences and Center for Data Science, as well as an Amazon Scholar, Andrew’s work spans model selection, uncertainty representation, Bayesian methods and time-series foundation models.

The session was geared toward Tower’s trading team members who are already familiar with recent advances in deep learning and LLMs. It explored several of the ideas shaping modern deep learning for time-series applications, including model construction, data selection, zero-shot forecasting, exogenous variables and uncertainty representation. Andrew also examined the role of synthetic data generation in improving forecasting performance, highlighting how synthetic data can help close the gap between large transformer-based models and smaller, more efficient neural networks.

Events like this reflect Tower’s broader commitment to creating an environment where learning, curiosity and technical rigor are part of the culture. By bringing in leading researchers from academia and industry, we aim to give our teams direct exposure to emerging ideas and encourage deeper thinking about the technologies shaping quantitative research and trading. That emphasis on continuous learning is a meaningful part of what makes Tower a place where talented people can keep growing throughout their careers.

We’re grateful to Andrew for joining us and sharing his perspective with our team.