The role :

You will join our ML team to turn research into working models for radio access networks. The focus is on predictive and learning-based methods for scheduling, beamforming, and interference management. You will build simulators, train and benchmark models, and help move the best ideas into production-grade pipelines. This is a strong fit if you want your research to ship and be measured against real system-level metrics.

What you'll do :

- Build and maintain Python system-level simulators for multi-cell massive-MIMO networks, including traffic, interference, and realistic channel dynamics.

- Develop deep learning models (graph neural networks, GRU/LSTM, Transformers) for time-series and spatio-temporal prediction in wireless systems.

- Integrate ML predictions into optimization pipelines such as coordinated beamforming, and quantify gains in sum rate, fairness, and cell-edge performance.

- Design reproducible experiments, ablation studies, and benchmarking workflows in PyTorch against classical and learned baselines.

- Explore reinforcement learning and Bayesian approaches for adaptive network decision-making.

- Partner with wireless researchers and software engineers to turn prototypes into reliable data and training pipelines.

- Write up results for internal reviews and, where appropriate, external publications.

What we're looking for :

- MSc (or equivalent) in Data Science, Machine Learning, Electrical Engineering, Statistics, or a related field.

- Strong Python and PyTorch skills, plus a solid grounding in deep learning, time-series modelling, and statistics.

- Hands-on experience with sequence models (RNN, LSTM, GRU, Transformers) and ideally graph neural networks.

- Experience with reproducible ML workflows, including Git, Linux, and experiment tracking.

- Working knowledge of SQL and data pipelines (ETL, validation, modelling).

- Curiosity and rigor : you test assumptions, report negative results honestly, and communicate clearly in English.

Nice to have :

- Exposure to wireless communications (5G/6G, MIMO, beamforming, scheduling).

- Experience with reinforcement learning (DQN, PPO, A2C) or Bayesian inference.

- Cloud data platform experience, such as Microsoft Fabric or Azure (DP-700 is a plus).

- A peer-reviewed publication or a strong research-style thesis.

- Familiarity with C++ or MATLAB for performance-critical or legacy code.

Compensation & benefits :

- Salary : [SEK range, confirm with HM], reviewed annually.

- 30 days paid vacation (Swedish standard of 25 plus company days), occupational pension, and parental leave top-up.

- Wellness allowance and a learning budget for conferences and courses.

- Relocation support and help with work permit or EU Blue Card applications where needed.

- Collective agreement coverage [confirm].