Thesis: Sovereign Voice AI for Sensitive Public-Sector Services

Public-sector organizations need voice services that can handle sensitive information without relying on external cloud infrastructure. This thesis explores whether locally deployed speech and language models can deliver the accuracy, speed, and reliability needed for real-time multilingual communication. The aim is to build and assess a reusable voice AI platform that keeps audio and text within infrastructure controlled by the organization.

The assignment

• Design and implement a modular research prototype using existing locally deployable models for speech recognition, language processing, translation, and speech synthesis, with reusable components accessible through APIs.
• Develop a reference application for multilingual conversations between citizens and public-sector staff, covering Swedish and two to three additional languages.
• Evaluate the prototype with synthetic conversations and volunteers, without using actual citizen data.
• Compare a limited selection of models and hardware configurations for accuracy, latency, computational demands, and reliability.
• Investigate how noise, accents, speaking speed, and specialist terminology affect transcription and translation.
• Explore methods for detecting consequential errors involving names, numbers, dates, monetary amounts, addresses, and negations, and for requesting clarification when interpretations are uncertain.
• Design for organizational control over sensitive audio, transcripts, and generated content, and deliver reusable components, a reference application, and a benchmark for future models.

Who we are looking for

• Studying towards a Bachelor or Master of Science in Computer Science or Engineering, with graduation planned for June 2027.
• Fluent in spoken and written Swedish and English.

What we offer

• An opportunity to develop a research prototype and reusable technical foundation for sensitive public-sector voice applications at Knightec Group.

Practical details

Based at the Sundsvall office. Start in January or March 2027. Apply through the careers page as soon as possible, and no later than 30 November 2026. Questions can be sent to johanna.edstrom@knightec.se.