Lead AI Research Scientist - S2S Weather Prediction

Lead AI Research Scientist - Subseasonal to Seasonal (S2S) Weather Prediction

What we offer

• A flexible schedule with up to 80% remote work, based in Barcelona.
• A permanent, full-time contract.
• €40,000 – €50,000 gross per year, depending on experience.
• Time and encouragement to publish your work and present it at scientific conferences.
• Ownership of a strategic research line from day one, working directly with our CTO.
• The opportunity to work with ground-breaking technology and create the future of subseasonal-to-seasonal prediction.
• High-impact, ambitious projects with real learning opportunities, in a company that respects your time and supports healthy work-life balance.

The role

We are looking for a PhD scientist to lead Nebbo's next – and most ferocious – research line: the design of a fully AI-based model for long-range weather prediction, covering horizons from day +15 to the coming months and years (subseasonal, seasonal and beyond).

This model will feed our existing products alongside the physics-based NWP models that already power them, and it is one of the most promising ways of enabling Nebbo's expansion into the energy trading sector. You will own this research line from day one, reporting directly to our CTO and working closely with our technical team.

Requirements

• PhD (already awarded) in physics, applied or computational mathematics, atmospheric, climate or Earth-system science, computer science, AI or engineering.
• Ideally, 2 years of post-PhD experience (postdoc or industry). Your PhD and/or your most recent work should revolve around ML/DL models for weather or climate prediction at long ranges – subseasonal, seasonal or longer.
• Proficiency in Python and deep learning frameworks (PyTorch or similar).
• Proven experience developing DL architectures for large-scale weather forecasting (graph-based networks, physics-informed ML, transformers, diffusion models…), and hands-on knowledge of state-of-the-art AI weather models such as GraphCast, GenCast, AIFS, FuXi-S2S or Pangu-Weather.
• Experience identifying and analysing predictability sources and modes of variability in the Earth system.
• Experience pre-training models on reanalysis data and working with the Earth-science data stack: ERA5, S2S/C3S hindcasts, xarray, zarr, dask, NetCDF/GRIB.
• Experience training and optimising models with experiment tracking tools (e.g.MLflow).
• Solid background in probabilistic forecasting and verification: ensembles or generative approaches, CRPS, reliability and skill scores against climatology.
• Experience in model evaluation and continuous validation & monitoring of deployed models.
• Fluent professional English, and the ability to explain complex results to customers and non-expert stakeholders.
• The right to work in Spain (or the ability to obtain it on your own), and being based in – or willing to relocate to – the Barcelona area.

Start date and funding

Hiring for this position is contingent upon Nebbo securing public funding through the Neotec (CDTI) and/or Torres Quevedo (Spanish State Research Agency, AEI) programmes.

• Neotec: a decision is expected in October–November 2026. If Nebbo receives this funding, you could join us at that point.
• Torres Quevedo: otherwise, the start date will depend on the Torres Quevedo resolution, expected in the second quarter of 2027. If you are selected, we will propose you as Nebbo

What you will build

- Phase 1 – Regional S2S model. Develop an AI model for subseasonal forecasting at country or electricity-market bidding-zone level, for a small number of regions.

- Phase 2 – Global S2S engine. Research and develop a worldwide AI-based engine for subseasonal-to-seasonal predictions.

Your responsibilities

• Design, develop and train deep learning architectures for large-scale weather forecasting (graph-based networks, transformers, generative and physics-informed ML).
• Identify and analyse sources of predictability and modes of variability in the Earth system (e.g. ENSO, MJO, NAO, the stratospheric polar vortex, soil moisture, sea ice) and exploit them in the models.
• Pre-train models on reanalysis data (e.g. ERA5) and fine-tune them with hindcasts and observations.
• Train and optimise models with a rigorous, reproducible workflow using experiment tracking tools (e.g. MLflow).
• Evaluate models against climatology and state-of-the-art dynamical forecasts, and set up continuous validation and monitoring of deployed models' performance.
• Work with our engineering team to bring models into production.
• Interact with customers and other stakeholders, turning their needs into forecast products.
• Represent Nebbo at fairs, workshops and scientific conferences.
• Help shape Nebbo's future in AI-based modelling and in the energy trading sector.

About Nebbo

Nebbo is a Barcelona-based startup, born in 2023 as a spin-off of the Vortex group, a world leader in modelled wind and weather data. We deliver subseasonal-to-seasonal (S2S) forecasts – from a few weeks to several months ahead – that combine physics-based numerical weather prediction (NWP) models with AI, helping companies in energy, agriculture and infrastructure plan with confidence. Our motto says it all: see further, plan better.

Nice to have

• Experience with cloud-based model training and deployment to a production model registry, ideally on Google Cloud Platform (GCP).
• Experience in the energy sector, ideally in renewable energies.
• Publications in peer-reviewed scientific and technical journals.
• Spanish and/or Catalan.