Om tjänsten
High Level Description
Recent advances in speech recognition, generative AI, and speech synthesis have created new opportunities for voice-enabled
services such as real-time translation, transcription, and conversational interfaces. However, organizations handling sensitive
information, particularly within the public sector, face challenges when using cloud-based AI services due to privacy and data
sovereignty requirements.
This thesis investigates whether locally hosted speech and language models can provide sufficient accuracy, performance, and
reliability for real-time voice applications involving sensitive information. The goal is to develop and evaluate a reusable
Sovereign Voice AI platform that processes audio and text entirely within infrastructure controlled by the organization.
Project Description
The student will design and implement a modular Sovereign Voice AI prototype using locally deployable AI models for speech
recognition, language processing, translation, and speech synthesis. The system should expose reusable components through APIs,
allowing it to support different applications.
The primary reference scenario will be real-time multilingual communication between citizens and public-sector employees, where
conversations may contain sensitive information such as personal details, financial information, or housing circumstances. The
prototype will be evaluated using synthetic conversations and volunteer participants rather than real citizen data.
A key focus will be developing mechanisms to identify potentially unreliable interpretations, particularly errors involving
critical information such as names, numbers, dates, monetary amounts, and addresses. The system should investigate ways to
highlight uncertainty and request clarification when necessary.
Different locally deployable models and hardware configurations will be compared in terms of transcription and translation
accuracy, latency, resource requirements, and reliability. The resulting solution should provide reusable components, a reference
application, and a benchmark for evaluating future voice AI models.
Who are we looking for?
Purpose
In this thesis, investigate the following questions:
1. Feasibility of locally hosted voice AI – To what extent can locally deployable speech and language models support real-time
multilingual communication under realistic hardware constraints?
2. Performance and efficiency – What trade-offs exist between model size, computational requirements, latency, and output
quality?
3. Accuracy and reliability – How do factors such as background noise, accents, speech rate, and specialized terminology affect
transcription and translation accuracy?
4. Consequential errors – How reliably can the system preserve critical information such as names, numbers, dates, monetary
amounts, and negations, and how can errors with significant consequences be identified?
5. Uncertainty and clarification – Can uncertainty-aware mechanisms identify potentially incorrect interpretations and request
human confirmation to reduce consequential errors without unnecessarily slowing down conversations?
6. Data sovereignty – How can the system be designed to ensure that sensitive audio, transcripts, and generated information
remain within infrastructure controlled by the organization?
The study will focus on existing locally deployable models rather than training new models from scratch. A reasonable scope is
Swedish and two to three additional languages, using one reference application and a limited selection of models.
The implementation is intended as a research prototype and reusable technical foundation rather than a production-ready solution.
An Exciting Journey with Knightec Group
Semcon and Knightec have joined forces as Knightec Group. Together, we are Northern Europe’s leading strategic partner in product
and digital service development. With a unique combination of cross-functional expertise and a holistic business understanding, we
help our clients realize their strategies – from idea to complete solution.
Practical Information
This is a thesis position, located at our office in Sundsvall. Start date January or March 2027.
Please submit your application as soon as possible, but no later than 2026-11-30. If you have any questions, you are welcome to
contact johanna.edstrom@knightec.se. Note that due to GDPR, we only accept applications through our careers page.