Senior Data Scientist - Foundation Model - AI Factory

BCO.BILBAO VIZCAYA ARGENTARIA; Madrid; Presencial; EUR 90.000 - 130.000; Jornada completa

Descripción de la vacante

BBVA in Madrid is seeking a Data Scientist with strong expertise in designing, training, adapting, and evaluating large-scale AI models. This senior role covers model architecture, training objectives, data preparation, distributed training, fine-tuning, evaluation, and production deployment, collaborating with engineers, researchers, data specialists, platform teams, and business stakeholders.

You will work across structured and unstructured data, investigate convergence and performance issues,

Formación

• 5+ years in leadership roles across technology, analytics, or AI-driven organisations.
• Experience with distributed compute and orchestration technologies.
• Independently own complex technical problems from investigation through implementation.
• Hands-on experience designing, training, or adapting large-scale models in research or production settings.
• Familiarity with financial datasets such as payments, transactions, credit information, risk indicators, or regulatory data.
• Knowledge of sequential modelling, GNN, LLM pre-training, architecture selection, data pipelines, distributed execution, optimization, and training stability.

Responsabilidades

• Define approaches for filtering, sampling, balancing, sequencing, and combining data from different banking domains.
• Design tokenisation, feature representation, masking, and sequence-generation strategies for financial information.
• Address characteristics of sensitive, sparse, imbalanced, and temporally ordered data.
• Detect and mitigate information leakage, historical bias, missing values, geographic variation, and changes in behaviour over time.
• Design, implement, and train models adapted to financial and banking information.
• Contribute to selection of architectures, objectives, training strategies, and representations.
• Work with structured and unstructured data sources.
• Investigate and resolve convergence, training stability, data quality, memory usage, and computational performance.
• Design reproducible experiments and evaluate alternatives with evidence and defined criteria.
• Develop adaptation strategies (continued pre-training, supervised fine-tuning, parameter-efficient methods) for financial use cases.

Conocimientos

Python

PyTorch

Transformer models

Distributed training

Data science

Experiment tracking

Memory optimization

Distributed compute

Educación

Bachelor's or Master’s in Computer Science / AI / Data Science

Herramientas

Docker

Kubernetes

Descripción del empleo

About the job

BBVA is developing state-of-the-art capabilities designed around the specific characteristics of financial services. As part of this ambition, we are exploring foundation models that can learn from the different forms of information generated across banking, including transactional sequences, product interactions, operational events… We are looking for a Data Scientist with strong expertise in the design, training, adaptation, and evaluation of large-scale AI models. This is a senior position for someone who can address difficult technical problems, contribute specialist knowledge, and take end-to-end ownership of important areas of the foundation model lifecycle. The role will contribute to decisions involving model architectures, training objectives, data preparation, distributed training, fine-tuning, evaluation, and production deployment. The successful candidate will work closely with other engineers, researchers, data specialists, platform teams, and business stakeholders, using their expertise to improve technical decisions and raise the overall quality of the initiative.

Key Responsibilities

Foundation Model Development

• Define approaches for filtering, sampling, balancing, sequencing, and combining data from different banking domains.
• Design tokenisation, feature representation, masking, and sequence-generation strategies suited to financial information.
• Address the characteristics of sensitive, sparse, imbalanced, and temporally ordered data.
• Detect and mitigate potential issues such as information leakage, historical bias, missing values, geographic variation, and changes in behaviour over time.
• Design, implement, and train models adapted to financial and banking information.
• Contribute expert judgement to the selection of architectures, learning objectives, training strategies, and customer and event representations.
• Work with structured and unstructured sources, including text, tabular datasets, transactions, event sequences.
• Investigate and resolve problems related to convergence, training stability, data quality, memory usage, and computational performance.
• Design reproducible experiments and evaluate technical alternatives using clear evidence and well-defined criteria.
• Develop and compare adaptation strategies, including continued pre‑training, supervised fine‑tuning, parameter‑efficient methods, and task‑specific modelling approaches, selecting the most appropriate technique for each financial use case.

Training Infrastructure and scale

• Work with platform and engineering teams to define, evolve, and use the technical environments required for reliable large‑scale model training.
• Improve accelerator utilisation through profiling, parallelism, scheduling, memory optimisation, and performance tuning.
• Monitor long-running experiments and implement mechanisms that improve their reliability, recoverability, and auditability.
• Evaluate model performance across countries, customer groups, products, time periods, and changing economic conditions.
• Support the transition of successful experiments into robust production implementations.

Ways of Working

• Work closely with researchers, engineers, data scientists, architects, product teams, and control functions across BBVA.
• Share reusable methods, tools, documentation, and engineering practices with the wider AI and data community.
• Explain model behaviour, technical trade‑offs, risks, and limitations to both technical and non‑technical stakeholders.
• Help teams understand where foundation models can provide value and where alternative approaches may be more suitable.
• Participate in internal technical forums and, where appropriate, external research or industry activities.
• Ensure consistent application of WoW (Ways of Working) guidelines for models and engines across all products.

Required Qualifications

Experience

• 5+ years of experience in leadership roles across technology, analytics, or AI‑driven organisations.
• Experience with distributed compute and orchestration technologies.
• Demonstrated ability to independently own complex technical problems from investigation through implementation.
• Substantial hands‑on experience designing, training, or adapting large‑scale models in research or production settings.
• Preferred familiarity with financial datasets such as payments, transactions, credit information, risk indicators, regulatory documents, or financial crime data.
• Practical knowledge of sequential modelling, GNN, LLM pre‑training, including architecture selection, data pipelines, distributed execution, optimisation, and training stability.

Skills

• Advanced programming ability in Python and strong software engineering practices.
• Deep understanding of modern machine learning and deep learning techniques.
• Strong practical experience with PyTorch or an equivalent deep learning framework.
• Detailed knowledge of transformer architectures, representation learning, optimisation, and large‑scale model training.
• Ability to implement and troubleshoot distributed training across multiple accelerators or compute nodes.
• Familiarity with parallel training, mixed precision, checkpointing, experiment tracking, memory optimisation, and performance profiling.
• Understanding of model evaluation beyond aggregate accuracy, including robustness, fairness, calibration, explainability, and operational reliability.
• Ability to work with sensitive, complex, imbalanced, and temporally structured datasets.
• Desired experience with full fine‑tuning, supervised fine‑tuning, LoRA, or other parameter‑efficient adaptation methods.

Education

Bachelor’s or Master’s degree in Computer Science, Artificial Intelligence, Data Science, Mathematics, Statistics, Physics, Engineering, or another relevant quantitative discipline. A PhD in a related field would be an advantage but is not required when the candidate can demonstrate equivalent technical expertise and practical experience.

Why Join Us

This is a unique opportunity to shape how AI products are built, governed, and delivered at scale within a global financial institution. You will work at the intersection of technology, governance, and innovation, driving initiatives that directly impact millions of users and position BBVA as a leader in AI development.

Skills

Client Orientation, Empathy, Ethics, Innovation, Proactive Thinking

Company Culture

We are more than 121,000 colleagues across 25 countries, working in multidisciplinary teams where we understand the importance of work‑life balance. We support our clients in the energy transition and are committed to inclusive growth. We are pioneers in adopting disruptive technologies that will shape the financial industry. Dare to define the future of banking!