Applied Data Scientist (Fintech)

TL;DR

Applied Data Scientist (PyTorch/Python): Building and optimizing ML models for document classification and ledger prediction in the accounting domain with an accent on production-grade modeling and LLM fine-tuning. Focus on implementing transformer models, designing evaluation pipelines, and ensuring model stability via drift detection.

Location: Hybrid in Antwerpen (Belgium) or Barcelona (Spain)

Company

International IT scale-up supporting accountants and entrepreneurs in handling their accounting more efficiently through innovative software solutions.

What you will do

• Train and evaluate classic ML (XGBoost, scikit-learn) and transformer models for document classification and ledger prediction.
• Perform model training and fine-tuning of LLMs using PyTorch and HuggingFace.
• Own feature engineering, model evaluation, and retrieval strategies for vector search.
• Maintain production quality by implementing drift detection, evaluation pipelines, and managing the MLflow model registry.
• Develop a deep understanding of accounting and financial-document structures to inform modeling decisions.

Requirements

• Strong applied ML background with proficiency in both classic ML and transformer models.
• Hands-on experience with Python, PyTorch, scikit-learn, HuggingFace, Pandas, MLflow, Databricks, and Snowflake.
• Proven track record of shipping models that maintain performance in production environments.
• Ability to communicate and explain complex ML concepts in plain business language.
• Fluent English is a must.

Nice to have

• Exposure to the financial or accounting domain.
• Experience working with large document corpora.

Culture & Benefits

• Generous training budget for specialized certifications and courses available from day one.
• Ample vacation days with an option to purchase additional days.
• Workation policy allowing up to 5 weeks of remote work per year.
• Access to lifestyle coaches and well-being programs.
• Autonomy of an innovative scale-up backed by the stability of the Visma Group.