{"author":"varunkmohan","children":[{"author":"varunkmohan","children":[],"created_at":"2022-04-28T19:34:43.000Z","created_at_i":1651174483,"id":31197182,"options":[],"parent_id":31197171,"points":null,"story_id":31197171,"text":"Hi everyone! My name is Varun Mohan and I'm the CEO of Exafunction (we're hiring!). We're excited to share a bit about our platform, which we think will be the best way to run deep learning and GPU workloads in the cloud, starting with inference. We've already proven out our product by managing inference for some of the world's biggest GPU users, namely autonomous vehicle companies. You can read our Series A blog post here, where we briefly describe the tech: https://exafunction.com/blog/series-a-announcement

I'm happy to answer any questions and speak to any of the technical challenges we've had to tackle to make your remote GPU code feel effortlessly local.","title":null,"type":"comment","url":null},{"author":"nharada","children":[{"author":"pqn","children":[{"author":"nharada","children":[{"author":"pqn","children":[],"created_at":"2022-04-28T20:36:43.000Z","created_at_i":1651178203,"id":31197907,"options":[],"parent_id":31197634,"points":null,"story_id":31197171,"text":"That's a great point. We've been mostly outbound so far, but this will be a bigger issue for us moving forward. We're thinking about how to lower the barrier to entry for this -- for instance, we can try to publicly release some docs and let anyone try out the system in a local container.","title":null,"type":"comment","url":null}],"created_at":"2022-04-28T20:13:47.000Z","created_at_i":1651176827,"id":31197634,"options":[],"parent_id":31197547,"points":null,"story_id":31197171,"text":"Makes sense. I think it's less about speed of response and more that for many SWEs, we don't necessarily have the authority to reach out and create relationships between our companies and another vendor. I certainly am not going to bother trying to get that process started unless I can do a little legwork before involving other parties at my company.","title":null,"type":"comment","url":null}],"created_at":"2022-04-28T20:06:27.000Z","created_at_i":1651176387,"id":31197547,"options":[],"parent_id":31197259,"points":null,"story_id":31197171,"text":"SWE at Exafunction here! We're not ready to be self-serve, so the contact process is the most straightforward for us to work with companies right now. But we respond fast :)

As to the tech, we have APIs closely resembling common deep learning frameworks, so once you add our Python/C++ client locally, you can change a small amount of code to start remotely using GPUs. We also have the ability to handle arbitrary stateful CUDA code for more complex use cases. On the server side, you can deploy our work scheduler inside your own VPC, so we take over orchestration for you as well.

Our customers are currently confidential, but safe to say we've seen a 5-10x decrease in cloud costs (or equivalently, the ability to fit 5-10x larger workloads given a GPU quota). It really depends on the utilization of your current workload.","title":null,"type":"comment","url":null}],"created_at":"2022-04-28T19:40:59.000Z","created_at_i":1651174859,"id":31197259,"options":[],"parent_id":31197171,"points":null,"story_id":31197171,"text":"This is cool, but the overhead to learn more is pretty high. I have to "contact sales" to either get started or get a demo, there's no technical details about how it works (I assume that it's basically like GPU/deep learning lambda functions?), and there's no case studies or any performance details showing how much of an improvement in cost I can expect to get.","title":null,"type":"comment","url":null}],"created_at":"2022-04-28T19:33:47.000Z","created_at_i":1651174427,"id":31197171,"options":[],"parent_id":null,"points":8,"story_id":31197171,"text":null,"title":"Show HN: Exafunction, efficient deep learning at scale","type":"story","url":"https://exafunction.com"}