Master Thesis: Open Knowledge Format

Join our Team 
 About this opportunity:   Ericsson Customer Product Information (CPI), the documentation delivered to customers, must match the technical design, the CPI template requirements, and the applicable release and configuration. In an agentic CPI workflow, locating evidence in input documents and determining which CPI topics a change affects are of vital importance. This master’s thesis will investigate how metadata annotations, ontologies, and document templates can improve CPI generation and maintenance. The aim is to design and evaluate an AI-assisted prototype that connects annotated design information to CPI template requirements and dynamically assembles applicable content for a selected release and configuration. The solution should flag missing evidence, trace content to its sources, and identify affected topics when designs or template requirements change. Design knowledge will reside in a semantic layer combining a structured knowledge base with an ontology-based knowledge graph. Portable Open Knowledge Format (OKF) concept bundles, built and exchanged with Apache Ossie, will capture source knowledge, while the graph will link concepts, evidence, and CPI topics. Conventional Retrieval-Augmented Generation (RAG) feeds isolated text chunks to large language models (LLMs), so it breaks document semantics, overlooks dependencies between entities, and lags behind changes in siloed sources. The prototype’s drafts and updates will be compared with this baseline to evaluate factual accuracy, completeness, traceability, update correctness, and authoring and review effort.   What you will do: •    Review relevant literature and current approaches for documentation generation, knowledge representation, and graph-based RAG (GraphRAG). •    Study CPI template requirements and define annotations and an ontology linking design concepts, topics, releases, and configurations. •    Design and implement a FastAPI prototype that uses Apache Ossie to ingest design proposals, interface specifications, and schemas into validated OKF concept bundles, organised in a knowledge base and linked through an ontology-based knowledge graph. •    Implement GraphRAG, combining vector search with graph traversal, so that LLMs assemble applicable topics with traceable references and flag missing evidence. •    Identify the CPI topics affected when designs or template requirements change, regenerate them, and check that unaffected topics remain correct. •    Create an appropriately protected dataset, including controlled changes to designs and template requirements, and establish reference topics reviewed by experienced CPI authors. •    Compare results with the RAG baseline using measures such as claim accuracy, content coverage, citation validity, affected-topic precision and recall, correctness of regenerated topics, and authoring and review time. •    Analyse limitations, data-quality dependencies, and suitable human oversight for CPI authoring.   The skills you bring:   •    Master’s-level studies in computer science, software engineering, or a related field. •    Proficiency in Python, including asynchronous programming, and experience with LLM APIs, prompt engineering, and RAG. •    Familiarity with FastAPI or a similar web framework, PostgreSQL or vector databases, and Docker. •    Ability to work with structured and unstructured technical documentation. •    Experience with knowledge graphs, RDF/OWL ontologies, OKF, data pipelines, frontend tooling, or open-source practices is an advantage. •    Interest in empirical evaluation, technical documentation, and responsible AI. •    Strong analytical, problem-solving, and communication skills.   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