Kseniia Kuvshinova

Machine Learning Researcher

Sber AI Lab

Sber AI Lab is the artificial intelligence research and development unit of Sber, one of Russia's largest banking and technology groups.

Researched tabular and time-series ML, shipped FT-Transformer in LightAutoML, and published work on time-series foundation models and adversarial robustness.

• Built an in-house regression version of TabPFN for business time-series forecasting; it reduced MAE by 7% against the production econometric baseline. • Implemented FT-Transformer in the open-source AutoML library LightAutoML (1.5k GitHub stars, ∼1.9k PyPI downloads/month): PR #120 (+268 lines, 9 files) merged in Sep 2023 and still shipped as the fttransformer model. • Benchmarked state-of-the-art tabular deep learning (FT-Transformer, TabPFN, SAINT, TabNet, ResNet) on an internal credit-scoring early-warning task.

Data Science Intern

Raiffeisen Bank Russia, CIB Products Division

Raiffeisen Bank Russia was the Russian business of Raiffeisen Bank International, with retail and corporate banking services.

Built analytics and ML models for Raiffeisen Business Online, then reported client experience issues used to prioritise service development.

• Built the analytics for the Raiffeisen Business Online service from scratch (Hive, Spark, SQL; Tableau and Power BI dashboards) and ML models of the service’s maturity. • Identified client-experience issues and delivered the reports used to prioritise the service’s development.

PhD, Machine Learning

Mohamed bin Zayed University of Artificial Intelligence (MBZUAI)

MSc, Information Technology and Engineering

Skolkovo Institute of Science and Technology (Skoltech)

BA, Economics (Mathematical Economics and Econometrics)

HSE University