Geolava
ML Researcher, World Models
About this role
At Geolava, we bring AI to the physical world. We are building spatial intelligence systems by leveraging world models to simulate the evolution of the physical world over time. We continuously model real-world dynamics, simulate actions, and forecast their impact on assets. We are backed by top-tier VCs and have been revenue-generating from day one. This is your opportunity to join as a founding team member to help define the future of spatial intelligence from the ground up. We are looking for an ML Researcher to help build Geolava's world model for the built world: a system that perceives the current state of properties and infrastructure, learns how they change over time, and forecasts what comes next, whether the driver is time, a renovation, a new construction, or a hurricane. You will work from raw earth observation and property data all the way to a model that can reason about state, action, and future state, and you will decide how we measure whether it works. This is a research role on a small team, and you will own real problems end to end: framing the question, designing the experiment, running it on real data at scale, and writing up what you learned. It is also a startup. Research that works here does not sit in a paper. It ships as a capability inside the product that property investors, lenders, and infrastructure owners use to make decisions. Responsibilities (What You'll Do)
- Design, train, and evaluate world models of the built environment that predict how assets change over time and in response to interventions and external forces.
- Develop representation learning and predictive modeling approaches over multi-temporal aerial and satellite imagery, property records, and other physical and administrative signals.
- Build rigorous evaluation for a domain with no public benchmarks: baselines, held-out temporal splits, calibration, and tests that distinguish genuine forecasting skill from leakage.
- Run experiments on cloud GPUs with attention to throughput and cost, and turn results into clear, decision-ready writeups.
- Work with the AI and engineering teams to take successful research into production as capabilities in Geolava's platform.
- Stay current with the world-model, self-supervised learning, and generative modeling literature, and bring the ideas that matter into our work. Your Background
- World-model research experience. You have built or contributed to world models: learned dynamics models, latent predictive models, action-conditioned generation, or closely related work, in academia or industry. This is required.
- PhD. You hold a PhD in machine learning, computer vision, or a closely related field.
- Deep learning fluency. You are comfortable training large vision models, adapting foundation models efficiently, and diagnosing why a training run behaves the way it does.