Master Thesis - Code - Graph - Based Context - AI - Assisted Software Engineering in Radio Baseband L1

• Join our Team

About this opportunity:

AI coding assistants and agent frameworks are increasingly used to support software development, answering architecture questions, aiding analysis and troubleshooting, and assisting with code changes. Their usefulness depends less on the underlying model than on the context they are given about the specific system they work on. For a large, long-lived codebase such as Ericsson's Radio Baseband L1, feeding a model raw source files, or relying on it to parse the code on demand, scales poorly: relevant details are missed, structural relationships stay implicit, and cost and latency grow.

This thesis explores a code graph, a structured and queryable model of the source code where nodes represent code entities (functions, modules, components, interfaces, signals) and edges represent their relationships (calls, dependencies, communication, and execution order), as the foundation for building domain context that an AI system can query instead of reading everything. The work will survey the state of the art, design and implement a method for extracting context from real Radio Baseband L1 source code, and evaluate its effectiveness against alternative approaches.

The results will provide evidence-based guidance on whether and how source-derived context should be built and maintained in an industrial domain, directly informing how Ericsson builds AI support for product development.

What you will do:

• Study the background and state of the art on code representation for AI (code graphs, code property graphs, static analysis, structural indexing, and retrieval or embeddings over code).

• Analyse a bounded set of Radio Baseband L1 components to identify the code entities and relationships that matter for domain understanding, including execution order.

• Design and implement a method and prototype for extracting a code graph and building domain context from source.

• Define what "effective context" means and how to measure it, covering accuracy, coverage, freshness, and cost.

• Evaluate the code-graph approach against alternatives, such as live code parsing and text or embedding retrieval, on representative domain tasks.

• Draw conclusions and provide recommendations, including how to keep context fresh and consistent with a given code baseline.

The skills you bring:

• You are pursuing a Master's degree in Computer Science, Software Engineering, Artificial Intelligence, or a related field.

• You have a solid foundation in software architecture, data structures such as graphs, and ideally static analysis or compilers.

• You are comfortable working with Python. The domain code is largely C/C++.

• You are interested in program analysis, knowledge representation, and AI-assisted software engineering.

• You are analytical, systematic, and able to work independently on a complex problem.

• You can communicate technical findings clearly and draw practical conclusions from your evaluation.

• This thesis is suitable for 2 students and corresponds to 30 ECTS credits per student.

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: 791756