AI Engineer (Spatial Intelligence)
Job description
AI Engineer (Spatial Intelligence)
Edinburgh, Scotland, UK | Onsite | Full-Time, Permanent
A global technology company with one of the world’s largest R&D organisations is hiring an AI Engineer (Spatial Intelligence) for its Edinburgh research centre. The team is building the next generation of maps, powered entirely by edge AI. They are developing foundation models that run on consumer devices, understand the user’s physical surroundings and retrieve relevant local data, without relying on the cloud.
As AI Engineer (Spatial Intelligence), you will work where multimodal foundation models, geospatial data and on-device inference meet. That means adapting LLMs and vision-language models to hardware limits, turning raw geospatial data into structures an LLM can reason over, and building fast retrieval across distributed spatial data.
This is an applied research engineering role. You build working prototypes and demonstrators, and those results decide which spatial AI capabilities move towards consumer products. The role suits a Machine Learning Engineer, Applied ML Engineer, AI Research Engineer or Geospatial AI Engineer who wants to apply LLMs to the physical world rather than another text-only use case.
Key Responsibilities
- Fine-tune, compress and deploy multimodal foundation models (LLMs and VLMs) for consumer-grade edge hardware and specialised spatial use cases.
- Design and prototype pipelines that turn geospatial data into structured, LLM-ready representations for spatial reasoning.
- Architect high-performance vector search and embedding-based retrieval over distributed spatial and geographic data.
- Build data pipelines that add real-time semantic context to mapping systems, known as thematic mapping.
- Develop agent-based workflows that combine models, retrieval and spatial tools into end-to-end systems.
- Build proof-of-concept demonstrators that test new model capabilities, and turn the results into clear technical recommendations.
- Evaluate model accuracy, latency and memory trade-offs on device, and drive decisions on architecture and deployment.
Required Experience & Skills
Essential
- PhD in Computer Science, Robotics, Engineering or a related technical field
- At least 2 years' of hands-on experience (in parallel to doctorate or since) fine-tuning or building on LLMs, foundation models or multimodal architectures such as vision-language models, beyond calling an API.
- Practical experience with geospatial data (maps, location or Earth-observation data) and/or vector embeddings and semantic search, including retrieval-augmented generation (RAG).
- Strong Python skills, including building multi-step LLM or agent pipelines with orchestration frameworks such as LangGraph, AutoGen or Pydantic AI.
- A prototyping mindset: able to take an idea to a working demo quickly and judge whether it is viable.
Desirable
- Deploying models under local hardware limits using ONNX, TensorRT or CoreML, including quantisation and other on-device optimisation.
- Building high-performance vector databases or approximate nearest-neighbour (ANN) search at scale.
- Background in remote sensing, LiDAR or point clouds, OpenStreetMap or mapping and location products.
Recruiter’s Note
The strongest candidates will combine hands-on LLM or multimodal work with real spatial data, for example a geospatial or remote-sensing ML researcher moving into foundation models, or an LLM/RAG engineer with mapping or location experience. On-device deployment experience is a strong plus, but it is not expected from everyone.
By applying to this role you understand that we may collect your personal data and store and process it on our systems. For more information please see our Privacy Notice (https://eu-recruit.com/wp-content/uploads/2025/01/European-Tech-Recruit-Privacy-Notice-2025.pdf).