Master Thesis: Spec-to-Code AI Development

Join our Team

Can AI turn software specifications into production-ready code - and make the development loop faster, more consistent and easier to evolve?

About this opportunity

AI is rapidly changing how software is developed. But what if development starts not with code, but with a precise, machine-readable specification - and AI, generators and automated validation work together in a closed loop from requirements to implementation?

In this thesis, you will work at the forefront of modern software development in Ericsson Fintech, combining AI-assisted engineering with real, large-scale financial technology. From Karlskrona, we develop technology that serves more than 600 million registered wallet users globally. You will work alongside some of the strongest software developers in the region and gain hands-on experience with how AI, specifications, automation and software engineering can come together in a production-oriented development environment.

The thesis focuses on Ericsson Offline Payments (EOP), where a specification-centric development approach is being explored. Requirements are versioned in Git and connected to artifacts such as user stories, OpenAPI specifications and PlantUML interaction flows. AI-assisted authoring is combined with reusable guardrails, deterministic generation and automated validation.

What you will do

Your challenge is to investigate how well this specification-to-code loop works in practice - and help make it better. You will combine software engineering, AI-assisted development and empirical evaluation.

Map where effort and lead time accumulate across specification authoring, code generation, hand-written integration, review and validation.

Compare specification-driven, AI-assisted development with a conventional code-first approach on comparable development tasks.

Experiment with improvements such as broader generator coverage, better AI guardrails, stronger validation gates or improved traceability.

Define meaningful measures for lead time, rework, consistency and defects, and use them to evaluate the approach.

Implement and demonstrate selected improvements in a realistic development environment.

Present your conclusions and recommendations to an Ericsson development team.

The exact focus can be shaped together with the supervisors. A primary contribution may emphasize either the comparison with code-first development or the design and evaluation of improvements to the Spec-to-Code pipeline.

Why this thesis is interesting

Work hands-on with generative AI in a real software engineering context - beyond ad-hoc prompting.

Explore the intersection of requirements engineering, AI-assisted development, APIs, code generation and automated quality gates.

Investigate a question with direct practical relevance: where does AI actually improve software development productivity and quality?

Contribute ideas and prototypes that can influence future ways of working in a large-scale Fintech R&D environment.

The skills you bring

Master-level studies in Computer Science, Software Engineering, Computer Engineering or a related field.

Solid programming and software engineering fundamentals.

Interest in generative AI, developer tooling, requirements engineering or automated software development.

Comfort working with Git and modern development workflows.

Curiosity and an experimental mindset - you enjoy measuring, comparing, building and improving.

Bonus: experience with REST/OpenAPI, PlantUML, code generation, CI/CD or AI coding assistants.

Practical information

Extent: 1-2 students, 30 ECTS credits (30 hp) each.

Location: Karlskrona, Sweden.

Preferred start: To be agreed.