About
Building Simmis
Simmis is a simulation platform for thinking through complex systems before you act on them — model your domain, branch the world, run scenarios, and keep what works. It's built on open-source, branchable infrastructure, and it's early. We're looking for collaborators and early users; the easiest way in is the open-source stack. For anything else, reach out to christian@weilbach.name.
Where Simmis comes from
Simmis grew out of a decade of work on memory and inference for intelligent systems. The premise is that good thinking is inference under a budget: many small, cheap tries beat one expensive perfect answer. That only works if trying another version is cheap — so the foundation is Datahike and the replikativ stack, where data, computation, and reasoning all branch like git. Simmis is the platform where humans and AI agents meet that substrate at one desk.
Who's building this
Christian Weilbach — Founder
I'm a researcher and systems builder focused on persistent memory and runtimes for intelligent systems. My PhD at the University of British Columbia was in Structured Amortized Variational Inference — building inference systems that accumulate evidence and maintain structured beliefs over time. That research directly shapes Simmis: data that versions, forks, and updates beliefs rather than overwriting state.
Before UBC I completed a Master's at Heidelberg University in Dictionary Learning with Bayesian GANs, and a Bachelor's implementing online learning for Boltzmann machines on spiking neuron models. I also spent several years in Philosophy and Cultural Anthropology — a different lens on how systems encode meaning.
On the research side: papers at ICML, NeurIPS, AISTATS, and TMLR; a Google grant for Graphically Structured Diffusion Models; talks at Google DeepMind London, OpenAI San Francisco, and MILA Montreal. The through-line of all of it: how do you build systems that accumulate knowledge, reason with their own history, and stay inspectable over time? Datahike was part of the answer; Simmis is the next part.
Join us
Build with us
The whole stack is open source and independently useful — Spindel, Raster, Ansatz, Dvergr, and the Datahike data family. Whether you're into probabilistic programming, reactive runtimes, GPU compute, verification, or agentic systems, there's room to contribute.
Work with us
Simmis itself is in early access. If the vision resonates — as a user, partner, or collaborator — get in touch and we'll find a way to work together.
Email: christian@weilbach.name