Joris Postmus
AI Researcher & Developer
Amsterdam, Netherlands
About
I think AI is the most significant human invention of our time, and as someone in the field, I feel a real responsibility to help make sure it develops in ways that are genuinely good.
This is what originally got me into AI safety. I previously co-founded AISIG, which started as a student initiative in Groningen and later became Safe AI Netherlands (SAIN), the national AI safety initiative in the Netherlands. I also do research on making language models more interpretable and steerable, including work on controlling model behavior presented at a NeurIPS workshop.
Over the years, I've noticed that the problems I care about most deeply often trace back to the same human patterns. Growing disconnection and polarization, collapsing epistemics, and an inability to coordinate even when the stakes are existential. I think a lot of this comes down to ego, self-deception, a fundamental disconnect from how our own minds actually work, and a prisoner's dilemma playing out at a global scale. I also think AI, if developed and used properly, gives us a real chance to help people grow past those patterns, epistemically, cognitively, and spiritually, at a scale that was never possible before.
Given where my skills, personal interests, and sense of purpose point right now, Waking Up is where I want to focus. I've used the app for over five years, and the practice has deeply transformed how I think, experience, and relate to myself and others. In my current R&D role, I'm bringing what I've learned across AI safety research, engineering, and product development to what I believe is the most important mission I could be contributing to right now.
Experience
Waking Up (Contractor)
I've used Waking Up for over five years, and the practice has meaningfully changed my life, as it has for many others. In my current R&D role, I'm drawing on what I've learned across AI safety research, engineering, and product development to help extend that impact to more people. I believe this may be one of the most important things technology can help us do. The way we understand ourselves and relate to experience shapes how we live, make decisions, and treat one another, and I think it sits beneath many of the deepest problems we face.
I co-founded AISIG, which started as a student initiative in Groningen and later developed into Safe AI Netherlands (SAIN), the national AI safety initiative in the Netherlands, where I now serve on the Supervisory Board.
AI Safety Initiative Groningen (AISIG)
I co-founded AISIG as a student initiative in Groningen. Over three years, we grew it into a broader AI safety organization, organizing more than 20 events, running courses for over 60 students, and supporting research that reached NeurIPS and ICLR. That work eventually led to the creation of Safe AI Netherlands, the national AI safety initiative in the Netherlands. I continue to serve on AISIG's advisory board.
Yara AI
As the first hire, I led engineering for a personalized self-improvement platform developed with clinical experts. My work spanned full-stack development, AI safety, and product design.
TA for Introduction to Artificial Intelligence and Basic Scientific Skills. Mentored ~10 students per year through individual and group sessions. Separately selected as a Study Buddy, providing one-on-one academic coaching for students with ASD or ADHD under professional supervision.
Research
Steering Large Language Models using Conceptors: Improving Addition-Based Activation Engineering
NeurIPS 2024 · Workshop on Foundation Model Interventions
Joris Postmus, Steven Abreu
This work explores conceptors as a way to steer language model behavior. Rather than representing a target as a single point in activation space, conceptors represent patterns as ellipsoidal regions. In our experiments, this gave us more precise control across several steering tasks and allowed us to combine objectives using Boolean operations such as AND, OR, and NOT.
Education
BSc Artificial Intelligence
University of Groningen · 2020–2024
Final grade: 8.3/10 · Thesis: 9.5/10 (Conceptor steering for LLMs)
BSc Computing Science
University of Groningen · 2022–2025
I was enrolled in the BSc Computing Science alongside my AI degree and took a range of courses covering programming fundamentals, algorithms and data structures, object-oriented programming, information security, and information systems.
High School
Malvern Collegiate Institute, Toronto · 2016–2020
Gold medal in province-wide coding contest (Skills Ontario) · Highest mark in Computer Science · Avg. 96%
Building
I started programming when I was about ten, making video games to play with friends. That turned into web development, then software tools, then AI applications. It's always been my main creative outlet. Some of those early projects are still playable at jorispos.github.io.
More About My Work
What is activation engineering?+
Activation engineering steers the behavior of large language models by modifying their internal activations during inference. My research explores conceptors as an alternative to traditional vector-based approaches, allowing multiple steering objectives to be combined using Boolean operations such as AND, OR, and NOT.
What is conceptor steering?+
Conceptor steering is a method Steven Abreu and I developed for controlling large language model behavior. Unlike traditional activation engineering that uses single vectors, conceptors represent steering targets as ellipsoids in high-dimensional activation space. This allows steering objectives to be combined using Boolean algebra. We presented the work at the NeurIPS 2024 Workshop on Foundation Model Interventions.
What are AISIG and SAIN?+
AISIG is the AI Safety Initiative Groningen, a student-led organization I co-founded in 2022. We ran courses, workshops, events, and research projects to help people understand and work on AI risks. Its growth later led to the creation of Safe AI Netherlands (SAIN), a national organization with chapters in multiple Dutch cities. I remain involved with AISIG as an advisor and serve on SAIN's Supervisory Board.
Feel free to get in touch if you'd like to talk about AI, research, or something I've written.