Master thesis: Fine-Tuning Foundation Models for Energy-Efficient 5G Orchestration

Ericsson; Stockholms kommun; On-site; SEK 167,000 - 234,000; Full time

Job summary

Ericsson in Sweden invites you to join as a thesis student to explore cloud-native 5G NF forecasting, focusing on adapting time series foundation models to telemetry data for proactive energy-aware orchestration.

You will collect data from lab environments, evaluate models in zero-/few-shot setups, and potentially apply LoRA/adapters. Strong Python skills and familiarity with PyTorch and Hugging Face are valued, with international collaboration in Stockholm-based research.

Qualifications

• Background in machine learning, computer science, or data science.
• Strong understanding of deep learning and time series modelling; experience with PyTorch is advantageous.
• Proficiency in Python and familiarity with modern ML frameworks (Hugging Face, GluonTS, or similar).
• Interest in cloud-native systems (Kubernetes, microservices) and sustainable computing.
• Ability to conduct technical literature reviews, design experiments, analyse results, and document findings.

Responsibilities

• Investigate adaptation of time series foundation models to cloud-native NF telemetry metrics.
• Collect data from lab and testbed environments; evaluate models in zero-shot and few-shot settings.
• Explore parameter-efficient fine-tuning techniques (LoRA, adapters).
• Provide recommendations for integrating foundation model forecasting into energy-aware orchestration pipelines.

Skills

ML background

Deep learning

Time series modelling

PyTorch

Python

ML frameworks

English proficiency

Tools

Hugging Face

GluonTS

Job description

About the opportunity

Cloud-native 5G deployments run Network Functions (NFs) as Kubernetes-managed microservices on shared infrastructure, and as RAN components and core NFs migrate to centralised cloud environments, their energy consumption, resource utilisation, and traffic patterns become critical inputs to orchestration decisions such as scaling, scheduling, and workload consolidation. Current orchestration controllers act reactively — after load changes have already occurred — leading to over-provisioning and energy waste. Accurate short-to-medium horizon forecasts of energy usage, CPU/memory utilisation, and traffic volume at the NF and pod level are therefore essential for proactive, energy-aware management of cloud-native 5G networks.

What you will do

This thesis will investigate the adaptation of time series foundation models (e.g., TimesFM, TTM) to cloud-native 5G NF telemetry metrics through domain-specific fine-tuning, with the goal of producing accurate and calibrated forecasts that can serve as inputs to energy-efficient orchestration systems. The work will encompass data collection from lab and testbed environments, systematic evaluation of foundation models in zero-shot and few-shot settings, and potentially parameter-efficient fine-tuning techniques (LoRA, adapters). The expected outcome may include:

• a benchmarking study comparing time series foundation models against classical baselines on NF energy, resource usage, and/or traffic prediction;
• a fine-tuning methodology for adapting foundation models to cloud-native 5G NF telemetry metrics;
• an analysis of model generalisation across NF types, prediction targets, and training data regimes; and
• recommendations for integrating foundation model-based forecasting into energy-aware network orchestration pipelines.

The skills you bring

• Background in machine learning, computer science, data science, or a related field.
• Strong understanding of deep learning and time series modelling; experience with PyTorch is advantageous.
• Proficiency in Python and familiarity with modern ML frameworks (Hugging Face, GluonTS, or similar).
• Interest in cloud-native systems (Kubernetes, microservices) and sustainable computing.
• Ability to conduct technical literature reviews, design experiments, analyse results, and document findings.
• Strong analytical and problem-solving skills.
• Ability to work independently while communicating effectively in an international research environment.
• Good written and spoken English.

Why join Ericsson?

At Ericsson, you´ll have an outstanding opportunity. The chance to use your skills and imagination to push the boundaries of what´s possible. To build solutions never seen before to some of the world’s toughest problems. You´ll be challenged, but you won’t be alone. You´ll be joining a team of diverse innovators, all driven to go beyond the status quo to craft what comes next.

Primary country and city: Sweden (SE) || Stockholm

Req ID: 791022