DinoPlusAI: Latency Optimized AI Platform

Jay Hu, CEO From vehicles that understand their environment to voice-controlled assistants with an evolving personality, the technology landscape has witnessed smart inventions in the past few years. However, even though businesses have promising technologies such as AI at their disposal that power these functions, they continue to cripple with latency challenges. High latency not only prolongs data processing but also affects network performance, which remains a significant operational hurdle for any organization. This is where edge computing makes a difference. It ensures faster data processing, which remains crucial when AI is used to obtain real-time data. For instance, in a smart manufacturing unit where intelligent systems are employed to monitor the manufacturing process, even a minor delay in obtaining and processing data about system failure can be fatal.

A company that eliminates every such possibility in any given scenario is US-based DinoPlusAI. The company brings a unique AI platform—DinoEdge—that ensures real-time outcomes and empowers latency-sensitive applications across industries. “In need of a latency sensitive application, one needs to process data in the right moment at the right place. With edge on-premise and on the cloud, that’s what DinoPlusAI focuses on,” explains Jay Hu, CEO at DinoPlusAI. How data is transmitted and processed plays a significant role in increasing latency. Hence, the team at DinoPlusAI identifies and assesses the data flow and helps reduce latency, making changes to data transmission and processing as needed. This capability makes DinoPlusAI’s solution ideal for three huge market segments—smart industry, smart retail, and smart city.

In an industry with a multitude of AI platforms, DinoPlusAI’s DinoEdge stands out as a vertically integrated platform that serves as an appropriate solution for video analytics, speech recognition, and other processes; real-time data processing remains the key. DinoEdge is powered by FlextyX, an FPGA AI Inference Accelerator based on Xilinx Alveo U50 and U280 cards. FlextyX supports many AI models, such as convolutional neural networks (CNN), long short-term memory (LSTM), and BERT. The platform also features an open full-stack for AI (OSA), including a powerful AI inference engine hardware, driver, API, compiler, and optimization tools (AutoML, sparsity, and low-bit quantization) and support for open-source frameworks.

A Futuristic Solution

Introduced recently as the company’s key offering, DinoEdge is a futuristic solution that focuses on latency-sensitive edge-computing applications. Considering the clients’ varied business needs, the solution can be deployed on-premise or on the edge cloud.


We place our computing resource very close to the site, which makes data transmission real time

Featuring 5G powered edge computing and a new breed of virtual processing units that offload management functions from CPU with domain-specific acceleration, this solution is here to change the communication and computation paradigm.

DinoEdge generates the data on the devices such as the smartphone, car device, or camera. The data is then processed by a DinoEdge node, either on-premise or on the cloud, and after that sent to the hyperscale cloud for further analysis. This is the data transmit view that the firm abides by. Further, DinoPlusAI carries out network configurations in different ways as preferred by the client. The first approach involves leveraging the camera, smart speaker, or other sensors used by the clients to capture data and process it locally with the DinoEdge node located on-premise server. The results are then sent to the public cloud, where the company has its central controller, which is used to display the data and interpret the data through a dashboard. The central controller allows the users to control everything at a single spot irrespective of how many on-premise computing nodes they have. Another approach that DinoPlusAI adopts is leveraging edge computing wherein the data is sent from the sensor to the Edge cloud, which improves the process performance while also reducing latency to a great extent compared to a classic public cloud configuration.

The last and the most traditional way to network deployment is simply sending the raw sensor data to the public cloud and process it there. “We can provide different ways to handle the data. If the customer wants the data to be processed in the shortest time, it can be on-premise. Otherwise one can opt for the Edge Cloud. On the other hand, if the application is not time-sensitive, the customer can opt for the public cloud. These deployment choices coupled with our DinoEdge make it scalable and a preferred choice for businesses,” explains Hu. Another factor that makes the solution stand out is that the central controller can be used with AWS or any other platform. This proves beneficial when the customers need to move the data to their own data centers or another server.

There is no problem as long as the server has a connection with the computing node. DinoPlusAI even offers the capability to support multiple sites and several nodes as needed. If a client has one site in the US and another in a different country, the DinoPlusAI team can still assist them. “We place our computing resource very close to the site, which makes data transmission real time. Secondly, our computing resource itself is real-time optimized, and hence, we call it latency optimized. In a nutshell, we can process the data extremely fast, resulting in reduced network and computing latency,” elaborates Hu.
On the central controller side, DinoPlusAI team provides a slew of built-in post-processing application to interpret the data, partnering with developers to provide a marketplace, but also enables customers for custom made applications.

An Ideal Solution for Smart City Projects and Even Healthcare

DinoEdge has turned out to be the ideal solution for smart city projects owing to its ability to ensure real-time outcomes and low latency. In such projects, camera sensors and other monitoring tools are used extensively to ensure the utmost security. The team at DinoPlusAI appropriately understands the sense of speed and performance desired to fulfill these monitoring needs and addresses them efficiently with its platform. Other major industries that the company deals with include manufacturing, oil & gas, and retail.

“The company brings a unique AI platform—DinoEdge—that ensures real-time outcomes and empowers latency-sensitive applications across industries”

An apt example of how DinoEdge benefits customers is its usage in shopping malls. The camera sensors installed in the parking lots of these retail outlets continuously capture the footage used to monitor the parking area’s traffic condition. The data obtained from these cameras can be helpful in further making parking smarter. Hence, in any establishment that employs any sensors to gather relevant data in real-time, DinoEdge can work wonders.

DinoEdge’s relevance, however, is not confined to smart city projects or smart manufacturing. Considering the pandemic’s impact on the world, DinoPlusAI has ventured into the medical space to contain the spread of COVID-19. With the screening efficiency as high as 200 people a minute, the platform can be integrated with thermal infrared imaging cameras for body temperature screening in any establishment. The solution’s safety aspect also ensures additional security monitoring features such as social distancing, counting of people, intrusion detection, and so on. It’s a scalable solution that supports a multitude of cameras and varied applications leveraging AI algorithms.

Healthcare remains a critical operational area for DinoPlusAI at present, where it will continue to impact the future. Smart manufacturing and smart retail are the other sectors where the company seeks to expand going forward. “From the technological point of view, the organization will be migrating from the FPGA base to ASIC in the next 18 months. We are going to leverage the bandwidth and low latency from the 5G networks as well,” adds Patrick Ang, VP - business development, DinoPlusAI. For the DinoPlusAI team, it’s all about continuously growing with its customers and ensuring that they get access to a next-generation solution that reduces latency and, at the same time, provides throughput and power.

Company
DinoPlusAI

Headquarters
Fremont, CA

Management
Jay Hu, CEO

Description
The company brings a unique AI platform—DinoEdge—that ensures real-time outcomes and empowers latency-sensitive applications across industries. In an industry with a multitude of AI platforms, DinoPlusAI’s DinoEdge stands out as a vertically integrated platform that serves as an appropriate solution for video analytics, speech recognition, and other processes; real-time data processing remains the key. DinoEdge is powered by FlextyX, an FPGA AI Inference Accelerator based on Xilinx Alveo U50 and U280 cards. FlextyX supports many AI models, such as convolutional neural networks (CNN), long short-term memory (LSTM), and BERT

DinoPlusAI