How does graffiti intelligence build an intelligent solution platform connected to tens of millions of devices?

Textual research: Guan Xianhao

Written by: Ni Xianhao

Internet of Things is another information technology revolution after Internet and mobile Internet.

After more than two years of development, networking and remote control have become the standard configuration of intelligent hardware, and more and more products focus on voice interaction. The progress of artificial intelligence in speech recognition, NLP and image recognition has also contributed a lot to the development of new intelligent hardware.

Graffiti Intelligence, born in Hangzhou, is an intelligent platform focusing on the interactive connection between devices and voice, and is committed to providing one-stop product intelligence solutions for domestic and foreign customers.

Before creating graffiti intelligence, Wang worked in Alibaba and was the first person in charge of Alibaba Cloud. On 20 14, Wang He and Chen Liaohan (co-founder of graffiti intelligence) left Ali to set up graffiti intelligence.

In the first year of its establishment, the main work of graffiti intelligence was to complete team formation and product construction. Subsequently, through business dealings with customers in different industries and according to the application scenarios of different industries, graffiti intelligence summed up the standardization scheme, and then completed the platform construction and realized productization.

In March of 20 15, graffiti intelligence set up an AI team, and defined the positioning of AI+ manufacturing industry. Judging from the actual situation, it is currently concentrated in large/small household appliances, electricians, lighting, security and other industries.

By the end of 20 17, graffiti intelligence has completed the A and B rounds of financing, and the B round of financing has reached several hundred million yuan.

Module +APP OEM+ Cloud to Create Industry Standardization Solution

At present, the products provided by graffiti intelligence for enterprises are software and hardware packaging delivery solutions. The cloud can be used for life at one charge, and the hardware part is charged according to the number of devices.

Specifically, graffiti intelligence first accesses the traditional equipment of customers, and uses networking modules to complete networking, including wifi, Bluetooth, BLE Mesh, GPRS, Zigbee, etc., which is compatible with various protocols.

Secondly, according to the scene corresponding to customer requirements, such as voice control, embedded systems, sensors and AI related technologies (voice recognition, computer vision, etc.) are used. ) to complete product intelligence. Among them, the sensor embedded in the product constitutes a wireless sensor network (WSN), and the way to access the Internet is accomplished by deploying a protocol conversion gateway.

Figure graffiti intelligent business flow chart

After the hardware intelligence is completed, the sensors embedded in the product will collect data and upload them to the data center (PaaS platform), such as the number of devices, the operation data of devices, the user behavior data and so on.

The above accumulated data is conducive to graffiti intelligence to improve AI algorithm and improve products. At the same time, the user behavior analysis ability provided by the data center can also help customers make decisions.

As part of the overall solution, graffiti intelligence also provides customized apps for customers' own brands. Correspondingly, graffiti intelligence also provides customers with an operating platform for information interaction with product consumers. The specific ways of interaction are as follows: customers push brands, use skills and other information, as well as consumers' timely feedback on product opinions and suggestions.

There are advantages in technology and data accumulation.

Through the solutions provided, graffiti intelligence supports many application scenarios, such as voice control, temperature adjustment, electricity statistics, water quality exceeding the standard reminder, lighting brightness level and so on. Judging from the frequency of application scenarios, voice interaction is more for customers, with an average daily voice interaction of more than 6 million times.

In terms of products, networking and voice control are the standard in the current layout industry, and graffiti intelligence has been standardized for the universal needs of the industry, forming a SoC standardization scheme for the electrical industry, for example. With the accumulation of more and more industry cases, the productization rate tends to increase.

From the perspective of industry distribution, customers are mainly distributed in household appliances, small household appliances, electricians, lighting and security industries. Excluding the cloud charges and the higher charges for some customers to choose embedded AI, graffiti intelligence will control the smart charges for a single device at a low level, and it can be copied on a large scale by using the existing productization capabilities in the future.

On the customer side, the proportion of graffiti intelligent sales staff is not high. R&D accounts for more than 70% of the total number of team members, and the rest are marketing, administration and sales personnel. The sales mode adopts direct selling mode.

Technically, the graffiti intelligence team was born in Alibaba Cloud and has a deep understanding of AI, cloud computing and big data. From a practical point of view, graffiti intelligence has accumulated tens of millions of access devices, which means that the technical team has a strong ability to cope with high concurrency and high availability. In addition, the ability of graffiti intelligent building platform to realize productization, as well as the independent research and development of cloud and embedded AI, all reflect strong technical capabilities.

In terms of data, the massive data brought by the cumulative access of tens of millions of devices has brought great potential for improving graffiti intelligence in AI-related technologies, such as algorithms. At present, the main data collected are the number of devices, the operation data of devices, and the interaction data between users and devices. At present, more than 20 billion equipment requests are processed every day, and the accumulated data amount reaches 1PB.

These data can have a positive impact on the improvement of AI algorithm and user behavior analysis. The most obvious advantage is that the products that can help graffiti intelligence can be used in more application scenarios.

Recently, Ai analyzed and interviewed Wang, founder and CEO of Graffiti Intelligence, and now I will share some of the contents as follows.

Software and hardware package delivery, shortening the delivery cycle of intelligent hardware.

Love analysis: How many kinds of graffiti products can be divided into? What is the form of product delivery?

Wang: Because graffiti products involve many technical points and customer needs, they are actually provided to customers as a packaged solution.

From the operational point of view, it can be divided into five characteristics: one connection, short intelligent cycle of networking; The standard of single module and networking module is unified; One-APP, with its own APP customization; One-Operation, provided by the operation center; I-Data provided by the data center.

At present, the delivery form is software and hardware integrated package delivery.

Love analysis: What improvements do you think have been made after the completion of productization?

Wang: First of all, after platformization, the products have been standardized, and the stability and safety of the platform products we deliver to customers in various industries have been well guaranteed.

Second, improve the implementation cycle. After the customer puts forward the demand, graffiti intelligence can deliver the demo on the same day. After the customer passes the verification, mass production can be realized in three to four weeks.

Love analysis: What other functions does graffiti intelligence support besides control and networking functions?

Wang: It can be mainly divided into two categories: all our equipment supports intelligent voice function, and the identification of dangerous animals and dangerous behaviors, which may be similar to the services of companies specializing in voice recognition and image recognition, but with different emphasis.

Love analysis: which industries will be laid out in the future?

Wang: Mainly in household appliances, electricians, lighting equipment, security and other fields.

Love analysis: the strategic positioning of graffiti intelligence is "AI+ manufacturing". How to understand it?

Wang: From the product form, graffiti smart products belong to the Internet of Things (IoT). But it is more inclined to use AI technology to serve the manufacturing industry, that is, AIoT.

Internet of Things is another emerging field after PC Internet and mobile Internet. The technical support for the future development of Internet of Things is AI. At the same time, the development of AI is also inseparable from the data accumulation of the Internet of Things on various devices.

Compared with the application of Internet of Things in other industries, graffiti intelligence is more optimistic about the huge market scale of manufacturing industry, which is more helpful for the long-term development of graffiti intelligence.

The accumulated customers are over 10,000, and data accumulation is the key to the development of AIoT.

Love Analysis: What is the graffiti intelligent customer orientation?

Wang: At present, graffiti has no scale requirements for customers. As long as customers have demand, we will undertake business. Of course, with the development of the market, it is not ruled out that in the future, it will be differentiated according to customer needs and positioned to specific types of customers.

Love analysis: what are the decisive factors to ensure the delivery of the hardware part of the product?

Wang: It mainly depends on the ability to control the supply chain. Some hardware is outsourced to partners, and some graffiti intelligence is responsible.

Love analysis: what is the competitive advantage of graffiti intelligence accumulation and development so far?

Wang: The most important thing is the precipitation and accumulation of data. The technical development in the field of graffiti intelligence is different from the past, relying more on a large amount of data to temper, which is suitable for AI.

Love analysis: What is the charging mode of products?

Wang: At present, customers' products are charged according to the number of devices, while the use of the cloud is a one-time charge and lifelong use.

Love analysis: If the customer requests to add AI functions, such as voice recognition, how much does the price of a single device change?

Wang: It depends on the type of artificial intelligence. If it is cloud AI, charge it once and use it for life. If it is embedded AI, charging by device will cost much more than the implementation of cloud AI.

Love Analysis: How many clients are there now?

Wang: It has provided intelligent services to more than 65,438+00,000 customers, and the total number of connected devices has reached tens of millions.

Love analysis: what is the team structure? What about R&D personnel?

Wang: 70%, and the rest are platform operators and salespeople. There are not many salespeople. Future team expansion will also focus on research and development.

Love analysis: Is the sales model direct selling or channel?

Wang: Direct selling.

Love analysis: how to achieve so many customers with relatively little sales?

Wang: First of all, so far, we have almost no market budget, because graffiti-based products and platform services have a good customer reputation, and our partners will be willing to recommend more users to us. Secondly, because the online registration process of graffiti platform with low threshold also makes it more convenient for many users to use, and customers are more likely to succeed.

Love analysis: Now many companies are beginning to cut into the Internet of Things industry. What do you think are the barriers in this industry?

Wang: First of all, the algorithm of AI, the accumulated advantages of customers of market entrants, and the number of connected devices. The most important thing is data accumulation.

Love analysis: What is your judgment on the development trend and driving force of the Internet of Things industry?

Wang: The biggest technical support for the development of the Internet of Things industry is artificial intelligence. By combining AI with various fields, various IOT devices can be exported. Specifically, by mining application scenarios, AIoT products can be exported and industry changes can be promoted.

For example, in terms of voice, intelligent speakers can be output by intelligently transforming speakers, deploying sensors to capture information, and endowing analytical capabilities to handle demands. Such products even have the potential to become the entrance of social products, such as Echo. These are the industry changes brought about by AI as the driving force of the Internet of Things industry.