20 set - Turbigo
Klarna
What You Will Do You will work on some of the most technically interesting modelling problems in fintech, training large transformer-based models on long sequences of real-world transactional events. You will design tokenisation schemes for numerical, categorical, and temporal features, making deliberate decisions about vocabulary size, sequence length, and information compounding. Your work spans the full model lifecycle from data preparation through to production and you will translate research decisions into systems that set the direction for how machine learning operates across Klarna. You will be part of a small, high-ownership team where what you build has genuine impact on the products Klarna ships.
Who You Are
Deep understanding of transformer architectures and sequence modelling, with the ability to reason through architectural tradeoffs
Hands‑on experience designing tokenisation schemes for heterogeneous feature types such as numerical, categorical, and temporal
Proficiency in Python, PyTorch, SageMaker, and Airflow
Experience owning the full model lifecycle, from training through to serving in production
Comfortable working in a small, high-ownership team on open-ended technical problems
Nice to Have
Experience with Triton kernels or GPU-level optimisation
Broader ML background spanning areas beyond deep learning
Experience with large-scale transactional or financial data
Background in ML infrastructure or MLOps
#J-18808-Ljbffr
20 set - Treviso
Growtize Holding
20 set - Roma
Würth Italia
20 set - Parma
Business Integration Partners
20 set - Civitanova Marche
Grafton Recruitment