Research -- Wout Schellaert

Intro

I do AI-related consultancy work on request. I have have completed 10+ projects clients such as OpenAI, Samsung, and other high-profile tech companies, as well as for policy-bodies such as the OECD and the Joint Research Center of the European Commission. Usually I am contracted for AI evaluation-, safety-, or research oriented work, but I am open to anything were my expertise is relevant.

Depending on your needs, that expertise could cover, among other things, AI engineering (deep learning, language models, traditional ML), benchmark & test design, software development, project management, research, policy advice, data science, teaching, or giving talks on various AI topics.

If you need advice, a helping hand, or a project completed, feel free to contact me at wout@schellaert.org.

Public projects

2023	OpenAI Conducted red-teaming and general safety focused evaluations of GPT-4 before its public release. See the technical report.
2023	Samsung Supervised the capstone projects of 25 students for the Samsung AI Innovation Campus.
2023	OECD Designed (and executed) a method for the evaluation of three modern language models on the PISA test battery designed for global and standardised evaluation of the competencies of 15-year-olds w.r.t. mathematics, reading, and science.
2022	Joint Research Center of the European Commission Analysed dominant problems with AI evaluation in its relation to EU policy, laying the groundwork for a new research area on the predictability of AI validity. See my published papers for more info.