Our customer is a US-based company providing a no-code website building platform with a large user base.
Detailed information about the client cannot be disclosed under the provisions of the NDA.
Our client’s no-code website builder offers dozens of tools, including a building toolkit and multiple add-ons enriching the platform’s capabilities. Nonetheless, the client needed to stay on top of the latest industry trends and provide their users with advanced website building and content management experience. The major objective was creating instruments to quickly make changes to the website structure without requiring complex code changes and high levels of technical knowledge.
Our team had to implement, train and test the ML models to ensure that they meet the client’s expectations, can handle the platform’s large user base and provide effective results powered by OpenAI’s GPT models.
In general, our main tasks included:
To simplify the functionality for users, we formulated the main approaches and explored how to handle input information limitations. The next step was to look for data and approaches for auto-tagging and training. Our team prepared and collected data on the HTML/CSS/JS markup and generated pairs of text descriptions with the corresponding code. We paid close attention to results validation since we had to meet not only design capabilities but also support a business logic layer.
Our team had to overcome several technical challenges related to the interaction of the generated code with internal platform objects. Overall, the integration involved a significant amount of work on the back-end and front-end sides of the platform.
As a result, the implemented GPT code generator is able to generate and change the website code according to text queries entered in the query string. Another built-in plugin is a full-featured service that eliminates routine copywriting while automatically creating blog posts, product descriptions, study cases and huge topics based on the tags entered.
AI-powered code generation tool
Our team enhanced the client platform with a machine learning code generation tool. The tool utilizes OpenAI’s GPT-3 model to create code based on natural language input from the user.
Our developers tested multiple ML approaches for code generation and identified the best policy for model training. The AI engine was powered by the OpenAI platform, various code and visual template sources.
We implemented Azure for services and business logic, and Codex and GPT-3 models to develop a code- and content-generating plugin. The OpenAI solutions performed and produced the best results when our team tested them against other methodologies like CodeRL and Code T5.
AI advisor enables users with limited coding knowledge to create or change complex website functionality without having to write code from scratch. Users can simply input their desired website feature using natural language, and the tool will generate the necessary code for them, or offer design options depending on the website functionality.
Moreover, the feature can help reduce the number of errors in the generated code, resulting in a smoother website development process and better user experience for visitors.
GPT-based content generator
Innowise Group implemented the GPT-3-based plugin to help users generate high-quality content for websites built on the client’s platform. The GPT-3 model is capable of generating text that is almost indistinguishable from content written by a human. Thanks to generative models, users can create content for different scenarios and use cases.
We started by creating an API for the website builder to communicate with the GPT-3 model. Our specialists designed a user-friendly plugin interface that allows users to input a topic or keyword and receive relevant AI-generated content.
We trained the model on a large dataset of articles and blogs to ensure the accuracy and quality of copies. This helped the model to learn the subtleties of language considering website goals and target audience. The plugin can generate SEO-friendly texts and product descriptions that help websites rank higher in search engine results.
Our approach to the project was highly collaborative, working closely with the client to ensure that AI plugins meet their requirements for a simplified platform operation. We followed the SCRUM framework throughout the entire development process. Our work was divided into sprints, with each sprint lasting for two weeks after planning meetings with the customer. Our team demonstrated the completed work to the client and gathered feedback during bi-weekly sprint review meetings. We used JIRA as a project management tool, Confluence for documentation work and Google Chat for day-to-day communication.
Over a span of six months, the team successfully integrated the AI tools into the system. Currently, we are adding new features and training models on new datasets to maintain and improve the platform’s functionality.
The AI integration into the website building environment had a significant impact on our client’s business. The machine learning code generation tool helped to reduce the time required to develop custom modules and components by up to 60%, helping thousands of users avoid monotonous work and documentation exploration. This feature attracts users who don’t have the extensive technical knowledge to change the structure of sites by adjusting the code. Overall, the AI system integration has been a major success, with users reporting significant time savings and improved website functionality.
The GPT-based content generator helped to speed up content creation, which allows users to partly replace copywriters for creating short product descriptions or long topics on thematic websites. The plugin is now able to generate unique and relevant content for websites in a matter of minutes.
The integration of OpenAI models aided in enhancing the search engine optimization of websites by generating optimized meta descriptions and titles for every page. This resulted in a 17% increase in the website’s search engine rankings.
Having received and processed your request, we will get back to you shortly to detail your project needs and sign an NDA to ensure the confidentiality of information.
After examining requirements, our analysts and developers devise a project proposal with the scope of works, team size, time, and cost estimates.
We arrange a meeting with you to discuss the offer and come to an agreement.
We sign a contract and start working on your project as quickly as possible.