Tobias Norlund



Text generation (large scale Transformer-based language modelling), Text classification (multi-label, multi-class), Semantic similarity (document/word representation, search), Text clustering (topic modelling)

Data Science

Deep Learning (text, images, structured data), Recommendation systems (collaborative filtering, content based, hybrid), Probabilistic/Statistical models, A/B testing


Python (PyTorch, Tensorflow, numpy, scipy, pandas, scikit-learn, spaCy), Large-scale distributed computing (Tensorflow multi-gpu / tpu, Spark), Deploying Machine Learning to production (Docker, Kubernetes, Convox, AWS), Development tools (git, Docker, jupyter, linux), Frontend development (React, JavaScript), .NET (C#, ASP.NET), JVM (java, scala)


English (professional), Swedish (native)


Master of Science: Applied Physics and Electrical Engineering, Linköping University, Sweden

2010 - 2016
Master: Signal and Image Processing
Master's Thesis: The Use of Distributional Semantics in Text Classification Models

Exchange semester, ETH Zürich

Various courses in statistics and machine learning, including Introduction to Natural Language Processing, Learning and Intelligent Systems, Computational Statistics and Computational Intelligence Lab


Tobias Norlund, Agnes Stenbom. Building a Swedish Open-Domain Conversational Language Model. 23rd Nordic Conference on Computational Linguistics (NoDaLiDa), 2021

Tobias Norlund, David Nilsson, Magnus Sahlgren. Parameterized context windows in Random Indexing. ACL 2016 1st Workshop on Representation Learning for NLP

Work Experience

  • Industrial PhD candidate / Recorded Future / Sept, 2020 – Present

    As an industrial PhD student at Recorded Future, I’m conducting novel research within NLP. I am in particular interested in developing methods of grounding language in other data modalities such as visual.

  • Data Scientist / Schibsted Media Group / Mar, 2016 – Sept, 2020

    Schibsted is among the biggest media groups in the nordics. I worked as Data Scientist in Schibsted’s central Machine Learning / NLP team, on various ML/NLP related use cases including: 1) Media content enrichment: Providing tag suggestions to articles and videos 2) Message intent classification on Schibsted marketplaces 3) Product and category classification for price comparison site Prisjakt 4) Algorithms for optimizing and personalizing the frontpage news feed for Schibsted publishers 5) Robot journalism through natural language generation 6) Video recommendation systems.

  • Master’s Thesis / Gavagai / Sept, 2015 – Feb, 2016

    Researched novel ways of parameterizing pre-trained Random Indexing word embeddings, that can later be fine-tuned for downstream NLP tasks. Resulted in accepted workshop paper at ACL 2016, see Publications.

  • Software Engineer (internship) / SICK IVP / Jun, 2014 – Aug, 2014

    Worked with software development and in particular computer graphics and WebGL. Designed multi-touch gestures for 3D object manipulation on tablet devices. Hands-on experience with Google Web Toolkit, WebGL and Java. The work required thorough computer graphics and linear algebra skills as well as ability to quickly get into existing code bases.

  • CEO / / Jan, 2008 – present allows you to create a custom Do-It-Yourself (DYI) cross stitch embroidery kit, based on an image of your choosing. The service uses image processing and color quantization algorithms to transform images to embroideries with world class quality. Developed solely by myself, mainly during high school and early undergraduate years, the project has teached me valueable entrepreneurial as well as full-stack programming skills.