Language model engineer

You’re a machine learning engineer who has deep experience working with language models in production.

We host a large and growing collection of open source language models. You’ll innovate, develop, and maintain the systems and implementations that enable our users to build with them. You’ll do things like improve our inference and training stacks, implement new methods you’ve found in papers, sort out nasty CUDA errors, run fine-tuning experiments, and help build out new language model-related features.

We’re looking for the right person, not just someone who checks boxes. You don’t need to know everything and have done it all, but you might be a good fit if you:

• Write clear, idiomatic code, and you can build things end-to-end.
• Keep up with multiple facets of language model research.
• Are very comfortable with theory, but you’re happiest when you can get your hands dirty.
• Have deep experience serving generative language models in production.
• Have deep experience adapting pre-trained language models to meet downstream requirements, e.g. through SFT, RLHF, etc.
• Have deep experience with the full MLOps life-cycle for generative language models.
• Understand the tradeoffs between building it yourself and adopting a framework and you’re not dogmatic in either direction.
• You’re product oriented and you want to help build something beautiful rather than just throw models over the wall.
• You have excellent communication skills. We think most of being a programmer is not programming. We want you to be able to communicate complex topics clearly, write down your thinking, write good docs, etc.

Email us: jobs@replicate.com