AI Chatbot

The massive proliferation of ChatGPT use cases made the benefits of generative AI models obvious to wide public across the globe. For businesses, Large Language Models (LLM), with Retrieval Augmented Generation (RAG) architecture at the root, offer great opportunities to leverage their accumulated knowledge base with increased efficiency and comfort in the process of user communications.

Our conversational AI bot has been conceived and implemented in pursuit of adding enhanced user experience while browsing the abcloudz.com website in search of relevant information.  

The AI-powered bots engage machine learning and deep learning algorithms allowing them understand and satisfy user complex queries and sophisticated requests far better. Depending on an intended implementation, besides a website AI chatbot assistant, the solution architecture may be further developed to provide for other channels of user communication via phone, or messenger.

Platform
Web, Mobile (iOS, Android)
Devices
Desktop, mobile
Tech stack
Node.js, NestJS, ReactJS, PostgreSQL + PGVector, Redis, LangChain, OpenAI ChatGPT and ADA v2 Models
Industry
Consulting, E-Commerce, Education, Entertainment, Finance, Fitness, Healthcare, Public Administration, Trade
Scope of work

Software architecture and design, front end and back end development, UX/UI design, LLM pre-training and integration, deployment, Quality Assurance

Tags
ai chatgpt ml PostgreSQL

Summary

Large Language Models like ChatGPT gave way to the creation of a conversational user interface (CUI) which enables human-like interactions. With a profound experience of AI-based developments under the belt, we rose to a challenge to build an AI-powered customer service bot capable of understanding user queries and generating relevant outputs in a way that exceeds common expectations.  

In addition to improving user experience, we took advantage of employing the tool for our internal marketing research purposes by gearing the AI bot to the abcloudz.com website back end.  

Implementation of this project allowed the website users to enjoy an alternative powerful means of communication. 

Challenges

Despite the substantial hands-on experience in building versatile AI-driven apps, our dev team was new to development of bots based on conversational AI platforms. The team had to gather, process, and leverage volumes of technical information standing behind the development process. 

Another challenge was to create a powerful AI chatbot tool, not only equipped with an extensive knowledge base, but capable of learning from previous user interactions to provide a higher level of response relevancy, accuracy, and completeness. 

Solution

Even though we have consolidated a vast experience of versatile AI engagements, the project team had to carry out a great deal of in-depth research related to Natural Language Processing (NLP), Large Language Models (LLM) and existing conversational AI platforms. 

The development design phase started with elaboration of software architecture placing the Retrieval Augmented Generation (RAG) model into its foundation. RAG is an AI framework designed to retrieve relevant information from external sources beside the corpus of training knowledge. This extraordinary functionality allows large language models (LLMs) to generate the most accurate and up-to-date information for app users. 

To reach the goal, the dev team incorporated a set of task-specific elements in the AI Chatbot architecture, such as the LangChain framework, a vector storage built upon PostgreSQL with PGVector extension, and the OpenAI ChatGPT 3.5 Turbo and ADA v2 Large Language Models attached via API. 

When the basic development phase was over, the developers proceeded to the exciting but challenging stage of the project – the language model training, coaching, and fine-tuning. This part was followed up with extensive chatbot operation testing by QA and user teams. 

The AI Chatbot project has demonstrated that our AI engineers possess knowledge, skills and expertise to cope with highly challenging tasks in the generative AI domain and build AI-powered apps with a human interface. 

Results

In close collaboration of our teams, we successfully delivered a web-based smart chatbot solution powered by AI algorithms.  It is easy to embed into a web app and caters for both sides of software users: website visitors and chatbot operators. Website visitors get a 24/7 comfortable service of an intelligent, prompt online virtual assistant, while chatbot operators, besides eliminating pressure from their Customer Support personnel and encouraging user communication, receive valuable feedback and stats for marketing purposes.  

As a result, business customers can enjoy an out-of-the-box, scalable conversational AI solution that gives its users enhanced experience from their communication with the app. 

Due to its exclusive architecture built upon the Retrieval Augmented Generation (RAG) framework, the generative AI model can be trained to act in multiple roles: support engineer, sales manager, fitness coach, legal adviser, etc. It is applicable in practically every industry, from Trade and Finance to the entire public sector, especially Education, Healthcare, and Public Administration. Proven efficiency and applicability of the AI-powered bots across multiple industries promise very optimistic prospects for their fast adoption in the near future. 

What else we’ve got from this project? We expanded our expertise in building apps with an AI engine under the hood and added one more successful AI project to our corporate portfolio.

    Hi there! How can I help you?

    ABCAssist