At Scandinavian Airlines, I specialise in frontend development with a focus on post-travel customer engagement. I led the full migration of the SAS Customer Service portal to Next.js, rebuilding it from scratch to improve accessibility, scalability and UX, while ensuring uninterrupted customer case submissions. I architected the app using modular principles, introduced consistent and accessible UI/UX patterns and added key navigation and form features, aimed at modernising the portal and making it more stable. I also established quality guidelines, documentation and comprehensive tests to support onboarding and long-term maintainability. I owned all QA and pre-release testing and mentored teammates to ensure alignment with the app’s architecture and patterns. The migration resulted in a 100% success rate for case submissions, a major reduction in technical debt and faster feature delivery.

Currently, I maintain that codebase and I am also involved in frontend objectives, in both mono- and polyrepo cross-team projects, focused on digitalising and replacing legacy systems. I oversee adherence to patterns, code quality and accessibility and also support backend integrations (Express) and regularly implement new features and A/B tests for the customer service team.

My role at Vidispine was in a SaaS setting and my responsibilities regarded front-end objectives, but were occasionally backend-focused (backend built on .NET supported by a PostgreSQL database). I was involved in the development of a feature-rich, web-based (React/TypeScript) media ingestion tool used by broadcasting companies to schedule, record and transcode content. I also worked on other projects, where I implemented features in projects for clients in the medtech sector. Additionally, I was responsible for building reusable React components for a commercial component library (using Google's MUI as a base).

As a software developer at Storytel, I developed an end-to-end neural network pipeline, which was responsible for synthesising speech from text, for generation of long-form audiobook content. I was responsible for and led the entire process, working independently (within a Data Science team) for the best part of my tenure at the company. My responsibilities included:

• Delivering a fully functional MVP that included support for training style-controllable models for content production (English and Swedish languages)

• Designing methods for content generation, engineering and analyzing training data

• Evaluating all models with tests and mean opinion scores

• Working together with other developers when designing the frontend for the engine

• Collaborating with and directing voice actors during data collection

• Leading the recording sessions, directing voice actors when recording content to use in the data for the pipeline

The pipeline I developed was eventually successfully used to train the company’s first AI voice and produce the company's first fully neural network-based audiobook content.

As a NLP developer at ReadSpeaker, I transitioned from speech synthesis to broader software development on both in-house and client projects. My responsibilities included:

• Developing C++ pipelines for text normalization and disambiguation

• Supervising QA processes across multiple languages

• Training various NLP models, including spaCy-based parsers/taggers and grapheme-to-phoneme automata

• Engineering data for text-to-speech objectives using the models I trained

Completion of Master's Education

Thesis Project: Transfer Learning in Speech Synthesis - Exploring Pretrained Weights Adaptation and Usage of Speaker Embeddings in Neural End-to-End Speech Synthesis

The project explored the feasibility of predicting speech from text in pre-trained neural networks, where:

• available data was 80% less than required and unprocessed, so that pre-training the network on different data might be an option

• obtaining data was not feasible, so zero-shot prediction, based on speaker embeddings, using clustering and multi-voice recordings, might be an option

Results showed that both options are viable for when training data is not available, but that the network is not able to pick on all of the new voice's vocal characteristics (such as prosody). The training pipeline was a mix of pyTorch and TensorFlow and training was done on a GCP instance.

Concentration in Natural Language Processing (NLP)

Concentration in Classics and Literature