AI Customer Service

The best way for customers to access your information.

Overview:

I built a custom chatbot for my sourdough baking equipment site, Shana's Sourdough, leveraging modern AI techniques to create an AI customer service assistant for customer interaction on the website. This bespoke chatbot serves as a knowledgeable virtual assistant, addressing customer queries regarding product information, shipping details, and sourdough baking advice in real-time. By seamlessly integrating AI technology into the website, the chatbot enhances user experience and facilitates efficient customer service.

​

Have a chat with it, here at Shana's Sourdough.

​

Technologies Used:

• LangChain Framework: Employed for flexible LLM application development, enabling seamless integration of custom prompting and RAG strategies.

• OpenAI 3.5 Turbo Model: Powering the chatbot's conversational capabilities, providing accurate and contextually relevant responses to customer inquiries.

• Python: Utilized for data processing, model training, and backend development.

• React and CSS: Implemented for frontend UI design, ensuring an intuitive and visually appealing user interface.

• Vercel: Selected for frontend hosting, offering effortless deployment and serverless architecture.

• AWS Lambda Functions and DynamoDB: Utilized for backend development, enabling serverless execution of code and efficient data storage.

​

Project Timeline:

Data Collection and Pre-processing:

The project commenced with the compilation of a comprehensive dataset comprising real customer inquiries extracted from our email contact forms. This dataset, consisting of over 350 messages, gave a foundation for what the model would need to be capable of handling. Leveraging LLM functions I was able to use a model to automatically tag and categorize the messages, streamlining the data labeling phase.

​

Model Implementation:

With the categorized dataset in hand, I meticulously curated domain-specific information for the chatbot to use. Initially, I explored the possibility of utilizing a RAG strategy for context retrieval. However, given the concise nature of the information I created, I opted to embed the all necessary data directly into the model's prompt. This decision streamlined the implementation process and optimized the chatbot's performance for real-time interaction.

​

UI Development and Deployment:

To provide users with a seamless interface, I developed the frontend using React and CSS. Leveraging ChatGPT's guidance, I rapidly prototyped and refined the UI design, culminating in a visually engaging chat interface. For deployment, I leveraged Vercel's serverless platform, enabling hassle-free hosting. I was then able to iframe the application directly into our Shopify site.

​

Backend and Deployment:

For the backend, tasked with serving chat responses and preserving conversation history, I opted for a serverless approach. Leveraging AWS Lambda functions, I established a seamless infrastructure capable of executing code in response to RESTful endpoint requests. These functions were integrated with AWS's DynamoDB, facilitating efficient storage and retrieval of conversation history data.

​

​

Achievements:

Through meticulous planning and iterative development, I successfully deployed a custom chatbot that aids in customer service on the Shana's Sourdough website. This project not only allowed me to hone my skills in AI technologies but also highlighted the practical application of machine learning in enhancing user experience and streamlining business operations. Additionally, I gained valuable insights into the challenges of working with real-world datasets and the benefits of using serverless technologies. By leveraging a structured approach and continuous iteration, I was able to overcome challenges and deliver a functional and efficient solution that exceeded expectations.