Volcanic engine Xiang Liang: machine learning and intelligent recommendation platform multi cloud deployment solution officially released

2022 year 7 month 20 Japan , Volcano engine 2022 Force The prime mover conference was successfully held in Beijing Nuojin hotel . In the agenda of the morning ,《 Recommended system practice 》 The author of a Book 、 He is also the head of the volcanic engine machine learning system —— Xiang Liang , Expanded the topic as 《 to open up AI infrastructure ： Give Way AI You can get 》 Speech . In his speech, he announced on behalf of the volcanic engine ： The multi cloud deployment solution of machine learning and intelligent recommendation platform was officially released .

Xiang Liang attended the volcanic engine Force Prime mover assembly

Volcano engine intelligent recommendation platform is equipped with Tiktok 、 Watermelon Video 、 Today's headlines have the same recommendation ability 、 Algorithm accuracy and architecture stability of enterprise SaaS Software . Starting today , Volcano engine intelligent recommendation platform will serve a wider range of enterprise customers , For different cost requirements 、 Different levels of privacy security requirements , Optional 4 There are two ways to deploy ： Public cloud deployment 、VPC Deploy 、 exclusive AZ Deployment and Private cloud deployment . For example, small and medium-sized enterprises can choose to use public cloud SaaS To access the intelligent recommendation platform to save costs ; And Finance with high privacy requirements 、 Securities companies can choose private cloud deployment , Purchase software and hardware by yourself , Set up the computer room by yourself .

The following is the original text of Xiang Liang's speech intelligent recommendation platform ️

Next, let me introduce the intelligent recommendation platform . Our intelligent recommendation platforms include 5 The three main features , Among them, in real-time and large-scale , We fully learn from the practice of byte interior , It can realize second level real-time and super large-scale training of recommendation advertising model .

 

Volcanic engine intelligent recommendation platform 5 Big feature

Let me give you another detailed introduction 1、4、5 Three aspects . First, we provide end-to-end capabilities . Here, end-to-end refers to , You don't have to care too much about the details of the whole system , You just need to enter the user's behavior , Then you can get their recommendation results . The details in the middle are handled by the recommendation platform .

 

---

 

But often many customers are the biggest for end-to-end platforms Concern（ worry ） It just can't be customized , Therefore, we also aim at the three aspects that users most need customized development , It provides you with the ability of in-depth development . For example, first of all, in Feature Engineering , We can Support you to customize features , Define different characteristics for recommendation system and advertising system according to their own business .

 

Feature Engineering ： Support custom features

Secondly, we focus on the model development ability , Launched a rock （Monolith） The training framework of . You can use this training framework , Define your own network structure , Then we can fully mobilize thousands of CPU perhaps GPU To train your mission . Therefore, the development of this model can also be deeply open .

 

Model development ： Support model development under rock training framework , It also supports the selection of industry templates

 

Last , Our customers come from different industries , For example, there is e-commerce 、 Content 、 Advertisements, etc. , We also introduced the ability to customize templates . You can define your own template based on your own industry characteristics .

 

---

 

Finally, it echoes what I said earlier , Today, we officially released a variety of deployment solutions for the recommended platform and machine learning platform . Here we propose four different deployment methods , The lowest cost is the deployment of public cloud , The second is to realize tenant isolation VPC Deployment way , Again, if you have some unique needs for exclusive physical resources , We also offer exclusive AZ Solutions for . Finally, if some companies have a strong demand for privacy protection , We also support private cloud deployment . I believe you can deploy in many ways , You can always find a part that suits you .

 

Intelligent recommendation platform multi architecture deployment ： Public cloud deployment 、VPC Deploy 、 exclusive AZ Deployment and private cloud deployment

 

The above is all I share today , Thank you very much .

 

---

Scan the code and fill in the following form , Can get 【 to open up AI infrastructure ： Give Way AI Within reach 】 Full video and full PPT：