Doctoral Student in Urban Heat Prediction Using AI and Earth Observation at KTH

Application ends: September 24, 2026
Apply Now

Job Description

KTH’s Division of Robotics, Perception and Learning is recruiting a doctoral student to map urban air temperature with generative AI, using conditional diffusion models to fuse satellite, forecast and crowd-sourced weather-station data into high-resolution, probabilistic heat maps of cities. The project is funded by Digital Futures.

At a glance

  • Position: Doctoral student (1 post), Computer Science
  • Supervisor and contact: Associate Professor Josephine Sullivan (proposed supervisor), sullivan@kth.se
  • Collaborators: Prof. Yifang Ban (KTH Geoinformatics) and Sebastian Hafner (RISE, Kista)
  • Funding: Digital Futures
  • Reference: PA-2026-2917
  • Apply: 24 September 2026, midnight CET/CEST

About the doctoral project

  • Treat dense urban air temperature estimation as an inverse problem
  • Develop conditional diffusion models conditioned on large, heterogeneous and incomplete data
  • Fuse satellite imagery, weather forecasts and crowd-sourced station measurements into probabilistic temperature maps
  • Work across machine learning, computer vision and remote sensing on a pressing climate-adaptation problem

Swedish doctoral positions are salaried jobs rather than scholarships: you are employed by the university, take doctoral courses, do research with a supervisor, and may teach up to a fifth of the time.

What KTH is looking for

  • Basic eligibility for doctoral studies: a second-cycle (master’s) degree, or at least 240 higher education credits including 60 at second-cycle level, or equivalent knowledge
  • English equivalent to English B/6 (mandatory)
  • Practical proficiency in deep learning, with demonstrated competence in TensorFlow, PyTorch or JAX (a must)

Nice to have

  • GPU experimentation and cluster computing (for example Docker, Slurm)
  • An earlier specialisation in machine learning and experience with computer vision and/or remote sensing (highly desirable)

Pay and location

Monthly salary under KTH’s doctoral student salary agreement; the posting gives no figure. Employment is renewed in steps (first up to one year, then up to two years at a time) for a total of up to four years of full-time doctoral education. The position is in Stockholm, Sweden.

How to apply

Apply through the official KTH page. The closing date is 24 September 2026, midnight CET/CEST. Apply for the position and admission through KTH’s recruitment system using the link on the official posting. The application must include copies of diplomas and grades and proof of meeting the English requirement (copies of originals must be certified, with translations into English or Swedish where needed), a CV, an application letter of at most two pages on why you want to do research and how it relates to your studies and goals, and representative publications or technical reports (for long documents, an abstract plus a web link). Applications must arrive by midnight CET/CEST on the closing date.

See all our artificial intelligence jobs, or browse more research jobs on ResearchJobs.in.

Hiring institution: KTH Royal Institute of Technology

Official advertisement: kth.varbi.com

How to prepare for this application

  • Diffusion models: revise conditional diffusion and how guidance works.
  • Remote sensing: know land surface temperature products and why air temperature differs.
  • Uncertainty: think about evaluating probabilistic maps with sparse ground truth.
  • Code evidence: the selection weighs programming ability shown in completed projects; link them.
  • Deadline: it closes on 24 September 2026.

About KTH

KTH Royal Institute of Technology in Stockholm, founded in 1827, is Sweden's largest technical university. In Sweden, doctoral students are employed with a monthly salary under a collective agreement; the doctorate corresponds to four years of full-time study, and up to 20% of working time may go to teaching or other departmental duties.

We send one confirmation email first. Every alert has an unsubscribe link.