careers
we're a small team building the easiest way for engineers to train custom models. three kinds of people thrive here.
coder
you will be working on the underlying compute orchestration platform that powers our training stack. expect to spend your days on frontier distributed-systems problems and inference engineering - gpu schedulers, fault-tolerant training loops, low-latency serving.
email it to ????@castform.com. figure out the address:
lord_of_the_rings_the_fellowship_of_the_ring → VrdOsWfelVwshipO harry_potter_and_the_philosophers_stone → harryZottY_and_theZhilosophYs_stone reinforcement_learning_fine_tuning → ????
stuck? send it to jobs@castform.com instead - we'll still read it.
scientist
you will be creating training recipes that drive faster convergence and higher performance ceilings without sacrificing generalization. push on the hard parts: long-horizon credit assignment, exploration under sparse reward, and keeping models stable across thousands of steps.
send us a paper from the last six months you really enjoyed reading, and a few paragraphs on why - what surprised you, what you'd push on next.
send it to research@castform.com.
storyteller
help developers understand what is possible with post-training. build interesting demos, write tutorials and essays, ship the kind of content that makes an engineer want to train their first model on a saturday.
tell us about the coolest demo you'd want to build on the platform — what the model does, why a developer would care, and how you'd get them to try it.
send it to stories@castform.com.