Design × alignment. Interfaces as alignment surfaces — making AI legible, contestable, humane. Fellowship to full-time over 1.5 years.
PracticeSpace — co-founder
LLM simulations of therapy patients for trainees. Founded with NHS CBT practitioners.
Machine Learning
BSc + MSc in machine learning at Aalto University, Finland.
ACL Workshop 2025
adding soon
I care about AI being good for the people it touches. Two questions hold most of my attention:
1 What's in the latent space, and can we learn to read it and reach in?
2 What happens where a human and an AI meet, and how do we design that interface?
Both questions point to something far more complex, subtle, and fertile than first appears, rewarding a kind of close attention I'm drawn towards.
Ideas I'd love to see exist but won't get to. Take any of them, no credit needed.
A reader that shows the latent space of what you're reading
Therapy-sim for hard conversations you're about to have
An interface to "reach into" a model's reasoning and nudge it
A personal site that's a place, not a poster
Margin notes that argue back
A bookshelf that remembers why you loved each book
Updated June 2026, from Helsinki.
Reading
Joe Carlsmith's "Otherness and Control in the Age of AGI."
Building
Rethinking this very site — from a single-screen poster into a place with corners.
Leading
The AISC 2026 team.
co-founded the EA chapter at Aalto
mentored EA + AI safety groups via CEA