a
🚀 Prompt University is now live! Dive in and explore the future of Generative AI learning Learn more
🚀 Prompt University is now live! Dive in and explore the future of Generative AI learning Learn more
An industrial machine manufacturer needs a scalable, intelligent solution to assist technical support “hotliners” in customer troubleshooting. The client is looking for the right generative AI project to kickstart internal innovation and stay ahead of the curve. Alongside a dedicated NLP consulting mission on a separate but related project, this presents the company’s first foray into machine learning and generative AI, guided by Effixis.
The client sells a wide variety of complex industrial machines with technical manuals exceeding hundreds of pages, making it challenging to provide timely and thorough support. Other technical documentation is embedded in solution diagrams, tutorial videos, and other formats.
The client’s technical documentation covers a range of sources, including PDF manuals, tutorial videos, and solution diagrams. We developed a unified knowledge base to store this data in a special format for the assistant to access. Developing LLM-powered chatbot solutions presents additional hurdles to correct for hallucinations and false responses. This required careful system design and prompt engineering to minimize the likelihood of such issues. Additionally, we aim to deliver one-on-one training sessions to onboard the app users and outline the strengths and potential weaknesses of our solution.
We developed a state-of-the-art intelligent assistant based on LLM technology. The system functions primarily as a “smart chatbot”, with access to proprietary, domain-specific knowledge relevant to the client. It has the capability to semantically search through machine manuals and tutorial videos to accurately respond to user questions. In effect, the chatbot acts as an intelligent agent, autonomously searching and referencing the relevant sources as required by the user’s question. Importantly, the system is built with transparency in mind, making it clear which sources it uses to respond to a question and raising a warning if an answer cannot be found. Additionally, the user interface (UI) allows for additional control to filter by machine type, source, and more.
One exciting feature is the intelligent search through tutorial videos. By searching against the subtitle transcriptions, we can open and display videos at the exact timestamp matching the user’s query. Given the scope and breadth of tutorial videos, this can greatly simplify the searching and retrieval of relevant information.
Below, you can find examples that illustrate how our internal intelligent assistant operates and the tangible benefits it brings to the organization.
Azure
Python
Generative AI (LLM)
Pinecone Vector Database
In the context of the case study, the implementation of the intelligent assistant based on LLM technology yielded significant results, addressing the client’s challenges effectively:
Newsletter
By clicking, your are confirming that your agree with our Terms & Conditions
