Research Engineer

Responsibilities

• Train models with GRPODesign and iterate RL training environments for retrieval – unstructured, structured, web.
• Own the entire training pipeline: from training data curation to wandb.
• Run small and large model experiments – yolo runs encouraged.
• Work on next-generation vision-first embedding models.
• Lead discussions on research – reading group.
• Directly with the ex-research CEO.
• Future: Manage a team of research engineers.

Perks

• Non-bureaucratic compute approvals: If you want to train a model, you can. We've budgeted 100,000 H100 hours for this role. If things go well, this number will be higher.
• Work on frontier methods that scale. No weird old-school AI.
• Everyone on the team can code – this might change in the future of course.
• Competitive compensation with generous early-stage equity, full medical and vision.

Requirements

• Not afraid of formulas – a BSc/MSc/PhD is an indicator of this (but isn't the only one).
• Thinks they can learn anything in 2 weeks, but isn't arrogant about it.
• Prefers .py to .tex.
• Familiar with vLLM/SkyRL/Megatron/etc.
• Comfortable with torchrun/accelerate/multi-node training.
• Clever about getting the data needed – or synthetically generating it.
• Finds easy solutions to hard problems, but doesn't mind getting their hands dirty with PyTorch or CUDA.
• Publications are a plus, but being able to critically evaluate research is a must.
• Familiar with 'You and Your Research.'
• Understands what it takes to do significant work.
• Must articulate ideas well!
• A big part of making successful models is telling people about them. This includes writing docs and technical reports at the minimum – and jumping on podcasts at the extreme.

Things You Should Know

• Startup work is always intense and sometimes frustrating: The nature of working on novel ideas is that not all of them pan out. It can be that you put blood, sweat, and tears into a feature or model and it just ends up not working through no fault of your own.
• We might publish, but cannot guarantee that we will.
• The role is in-person only from our offices in SF and Zürich. If you do not have US work authorization, we can help with that.