Senior Data Scientist

About the role

We are expanding our Data Capture Research team in Prague to tackle sourcing decisions – which suppliers to invite, how to compare bids, and what constitutes a fair price. The role combines multimodal document understanding, recommendation, forecasting, game theory and combinatorial optimisation on a unique $10 trillion spend dataset.

Key responsibilities

• Own end‑to‑end sourcing research initiatives, from problem framing on real spend data to production deployment.
• Define training sets, labels and benchmarks that make sourcing problems learnable, deriving supervision from historical events.
• Design and run bold experiments with fast prototypes, honest baselines and offline evaluation that predicts online behaviour.
• Leverage Rossum’s proprietary T‑LLM foundations to extract signal from RFPs, specs, quotes and contracts.
• Ship models with production constraints in mind – inference cost, latency, robustness and safe failure handling.

Required profile

• 5+ years of applied ML, data science, ML engineering or quantitative research with productionised models.
• Strong Python skills and comfort working with large, messy datasets, including SQL.
• Depth in at least one modelling discipline and curiosity across others (deep learning, recommendation, forecasting, optimisation, causal inference, RL, mechanism design, LLMs).
• Scientific rigour in experiment design, metric validation and leakage prevention.
• Ownership mindset and ability to thrive in green‑field, ambiguous problem spaces.

Required skills

• Python
• SQL
• Deep learning frameworks (e.g., PyTorch, TensorFlow)
• Gradient boosting libraries (e.g., XGBoost, LightGBM)
• Optimization and operations‑research techniques
• Reinforcement learning / bandits
• LLM‑based systems

What we offer

• Access to a unique $10 trillion spend dataset and in‑house T‑LLM architectures.
• Full ownership of problems with direct impact on customers’ procurement decisions.
• State‑of‑the‑art compute resources, including high‑end GPUs and large‑memory clusters.
• 33 days of paid time off, parental leave, and a modern office with a terrace overlooking Prague Castle.