This job is no longer available
There are still lots of open positions. Let's find the one that's right for you.
We are now seeking a Senior Deep Learning Software Engineer for Recipe Pathfinding. NVIDIA is looking for experienced software engineers to help rethink and create SW systems to accelerate the discovery of new low-precision and sparsity recipes. A recipe defines which operators in an LLM are transformed into low-precision and/or sparsified variants, thereby unlocking efficiency gains. Recipes can be statically defined at model load time or dynamically adapting to a layer’s input distribution. Recipes can incorporate and compose algorithmic techniques like rotations or low-rank decompositions to tame aggressors like outliers. We are a team committed to developing next-generation SW to make use of novel HW features on Blackwell, Rubin, and beyond. The scope spans all phases of the LLM life cycle: pre training, post training, and generation. This is a coding-heavy role focused on infrastructure, tooling, and performance. Making these recipes fast is critical to minimize spend – each production run can easily cost millions of dollars. The candidate's work will directly support NVIDIA's internal SW systems for supporting recipe prototyping. Your work is a component of our SW productization story in libraries like Megatron-LM and Transformer Engine.
