AI Data Scientist

Data Scientist- AI

Barcelona, Spain

Hybrid- 2 days a week

We’re looking for a Data Scientist to help build and improve AI‑powered document processing solutions used across a large, complex organisation. You’ll work across LLMs, multimodal models, OCR, NLP and classical ML to turn messy, unstructured business documents into reliable structured data that drives real value.

In this role, you’ll design and evaluate approaches for document extraction, classification and validation — “Design, prototype and evaluate approaches for document extraction, classification, validation and transformation.” You’ll also work with highly variable formats — “Work with complex and variable formats, including multi‑page documents, tables, line items and inconsistent layouts.”

What you’ll do

• Develop and optimise prompts, structured outputs, validation logic and post‑processing components.
• Compare LLMs, multimodal models, OCR and deterministic approaches to select the right solution for each use case.
• Build representative datasets, define ground truth and establish robust evaluation metrics.
• Run experiments, analyse failure patterns and iterate to improve accuracy and reliability.
• Create reproducible Python workflows and contribute reusable components for extraction, validation and data prep.
• Collaborate with Product, AI Engineering and business teams to translate requirements into measurable data science problems.
• Communicate findings clearly to both technical and non‑technical audiences.

What you’ll bring

• 3+ years in data science, machine learning or applied AI.
• Hands‑on experience with LLMs, prompt engineering and experimentation.
• Strong NLP/ML fundamentals and the judgement to know when classical methods outperform LLMs.
• Solid statistics — enough to judge whether a performance change is real.
• Experience with unstructured or semi‑structured documents.
• Python skills that make your work reproducible and easy for engineers to productionise.
• Good coding practices: testing, version control, code review, reproducible experiments.
• Ability to explain technical trade‑offs clearly.
• Comfortable diving into business domains to understand what “correct extraction” actually means.

Nice to have

• Document intelligence / OCR (e.g., Azure Document Intelligence).
• Experience with managed AI platforms (Azure OpenAI, Amazon Bedrock).