Risk AI Data Scientist (Fintech)

TL;DR

Risk AI Data Scientist (LLM/Fintech): Integrating advanced AI capabilities into bank risk management with an accent on LLMs, NLP, and agentic workflows. Focus on designing the cognitive layer of risk environments, optimizing RAG pipelines, and fine-tuning open-weights models for production-ready compliance.

Location: Warszawa (Pańska 97), Poland

Salary: 13,000 – 22,000 PLN

Company

hirify.global is a global banking center providing risk identification, aggregation, and insight capabilities for the group across various risk domains.

What you will do

• Fine-tune open-weights models (e.g., Llama, Mistral) on GCP GPUs to handle banking risk management and credit policies.
• Architect Retrieval-Augmented Generation (RAG) systems for high-precision interaction with internal policy documents and regulations.
• Design and implement agentic AI workflows using LangChain/LangGraph to support multi-step risk management tasks.
• Build NLP pipelines to extract complex signals and features from unstructured data like PDFs and OCR outputs.
• Evaluate and monitor GenAI systems for hallucinations, quality, and drifting.
• Develop clean, modular Python code in Azure DevOps to ensure models are reproducible and deployment-ready.

Requirements

• Master’s degree in mathematics, economics, or an equivalent field.
• 7+ years of experience in risk management (risk modelling experience is a plus).
• Proficiency in Python, SQL, and PyTorch/TensorFlow.
• Experience with HuggingFace (Transformers, PEFT), LangChain/LlamaIndex, and Vector Stores (FAISS/Vertex Search).
• Hands-on experience with Google Cloud Platform (Vertex AI, Workbench) and Azure DevOps (Git, Pipelines).
• Proven track record with end-to-end pipelines from data collection to model deployment.

Nice to have

• Experience working within Agile/Scrum teams.
• Knowledge of AI risk governance.
• Experience with SAS.

Culture & Benefits

• Opportunity to work on the cognitive layer of a major bank's risk environment.
• Access to high-performance GCP GPU infrastructure for model fine-tuning.
• Integration into a global risk management architecture.