Speaker unverified: prediction

Transcript context

…Even in 2022, it's interesting. I would not say absolutely no, but cloud providers have some of those, but it's not first class citizen, right? At 2022, there's still like TensorFlow is massively in production. And this is all pre-gen AI, and PyTorch is kind of getting more and more adoption. But there's no PyTorch-first SaaS platform existing. At the same time, we are also a very pragmatic set of people. We really want to make sure from the get-go, we get really, really close to customers. We understand their use case, we understand their pain points, we understand the value we deliver to them. So we want to take a different approach instead of building a horizontal PyTorch cloud. We want to build a verticalized platform first. And then we talk with many customers. And interestingly, we started the company in September 2022, and in October, November, the OpenAI announced ChatGPT. And then boom, when we talked with many customers, they were like, can you help us work on the JNS aspect? So of course, there are some open source models. It's not as good at that time, but people are already putting a lot of attention there. Then we decided that if we're going to pick a vertical, we're going to pick JNI. The other reason is all JNI models are PyTorch models. So that's another reason. We believe that because of the nature of JNI, it's going to generate a lot of human consumable content. It will drive a lot of consumer, customer-developer-facing application and product innovation. Guaranteed. We're just at the beginning of this. Our prediction is for those kind of applications, the inference is much more important than training because inference scale is proportional to the up-limit award population. And training scale is proportional to the number of researchers. Of course, each training round could be very expensive. Although PyTorch supports both inference and training, we decided to laser focus on inference. So yeah, so that's how we got started. And we launched our public platform August last year. When we launched, it was a single product. It's a distributed inference engine with a simple API, open AI compatible API with many models. We started with LM and then we added a lot of models. Fast forward to now, we are a full platform with multiple product lines. So we love to kind of dive deep into what we offer. But that's a very fun journey in the past two years. What was the transition from you start to focus on PyTorch and people want to understand the framework, get it live. And now say maybe most people that use you don't even really know much about PyTorch at all. You know, they're just trying to consume a model. From a product perspective, like what were some of the decisions early on? Like right in October, November, you were just like, hey, most people just care about the model, not about the framework. We're going to make it super easy or was it more a gradual transition to the model library…

Stored transcript either side of the excerpt. The highlighted words are the published quote; the surrounding text is unedited source, never generated.