Master's Thesis: From Tokens to Actions: System-One Decision Models for Low-Latency Wireless Control

About Huawei Sweden R&D Founded in 1987, Huawei Technologies has grown into a global leader in telecommunications and network solutions, with innovation at the core of everything we do. Huawei established its first overseas R&D office in Sweden in 2000, and since then, we’ve expanded rapidly. Today, our R&D engineers are pushing boundaries in Stockholm, Gothenburg, and Lund, working on groundbreaking 6G wireless systems that shape tomorrow’s technology. Join us on our journey to develop the future of tech and be a part of a team that thrives on innovation, works hard to solve the industry’s biggest challenges, and collaborates with top engineers worldwide. If you’re ready to contribute to world-changing advancements, we want you on board! About the Project: System-One Models for Wireless Decision-Making Applications and Performance Evaluation This thesis evaluates System-One models — probabilistic, decision-native models that map an observed network state and a typed query directly to a calibrated distribution over a bounded action set, replacing token-by-token LLM generation — as the execution engine for the fast decision path of an agentic cellular modem. Expected Contributions: To formalize the decision structure of representative wireless control tasks as System One queries. To implement and evaluate system-one model calibration pipeline for those tasks using open wireless datasets or simulated state-action traces. To design and benchmark a confidence-threshold escalation policy against pure-LLM and pure-fast-path baselines. To characterize the quality–latency–calibration trade-offs across calibration levels and decision tasks. Timeline Given our current project schedules, we are looking for master’s students available to start in spring 2027. Qualifications & Skills Background : Master's student in Artificial Intelligence, Machine Learning, Computer Science, Software Engineering, or a related engineering field, with a strong interest in agentic AI system. Communication : Strong command of English, both spoken and written. Team & Independence : Capable of working effectively both in teams and independently. Commitment to Quality : High attention to detail and dedication to producing high-quality work. Desirable Experience : Programming experience in Python. Knowledge of large language models (LLMs), AI agent frameworks, prompt engineering, or reinforcement learning is an advantage. Are you ready to apply your skills in a real-world setting and help us lead the charge in tech innovation? Join Huawei Sweden R&D, and let’s make tech history together!