Kashif Rasul

Kashif Rasul is a machine-learning researcher whose work spans image benchmarks, NLP tools, probabilistic forecasting, and mathematical reasoning. He earned a mathematics PhD at Freie Universität Berlin in 2010 and worked in its Databases group and Zalando's Article Sales Forecasting group by 2016. He coauthored Fashion-MNIST in 2017 and FLAIR in 2019, then TimeGrad in 2021 and Lag-Llama in 2023. In 2024, he is credited as an author of NuminaMath 7B TIR, which wins the first AIMO Progress Prize. As of October 2026, he is a Principal Research Scientist working on deep learning, time-series forecasting, reinforcement learning, and high-performance computing.

Now at: Morgan Stanley and Hugging Face
Works: 43
Citations: 7,189
h-index: 8
Badges: Rising
Bridge

Publication numbers from OpenAlex, as of 22 October 2026.

Timeline

2010

A mathematics PhD

He earns a PhD in Mathematics at Freie Universität Berlin in 2010.
2016

Data scientist at Zalando

In 2016, he holds a position in Freie Universität Berlin's Databases group and works as a data scientist in Zalando's Article Sales Forecasting group.
2017

Fashion-MNIST dataset

With Han Xiao and Roland Vollgraf, he coauthors Fashion-MNIST in 2017, an image dataset for benchmarking machine-learning algorithms.
2019

A framework for NLP

He coauthors FLAIR in 2019, a framework for training and distributing sequence-labeling, text-classification and language models.
2021

TimeGrad diffusion model

In 2021, he coauthors TimeGrad, an autoregressive model that uses diffusion probabilistic models for multivariate time-series forecasting.
2023

Lag-Llama foundation model

In 2023, he coauthors Lag-Llama, a general-purpose foundation model for univariate probabilistic time-series forecasting.
2024

NuminaMath wins AIMO

In 2024, he is named as an author of NuminaMath 7B TIR; the model wins the first AIMO Progress Prize and solves 29 out of 50 problems on the private test set.
As of October 2026

Principal Research Scientist

He works as a Principal Research Scientist on deep learning, time-series forecasting, reinforcement learning and high-performance computing.