Stemar

Stemar was hired by a web development studio to construct an AI-powered knowledge base for a large industrial manufacturing environment. Selected for their cost-effective rates, excellent culture fit, and strong company values, Stemar meticulously crafted the knowledge base using a tailored Retrieval-Augmented Generation (RAG) approach, which significantly improved its efficiency. The system provides fast and reliable access to over 10,000 pages of technical documentation, enabling operators to retrieve pertinent information effortlessly.

The project involved a team of 2-5 employees from Stemar. They developed several key deliverables for the industrial manufacturing application. For instance, custom parser for technical documentation that preserved the structure of engineering tables, specifications, and maintenance instructions were created. They also incorporated a table-aware chunking logic to ensure data precision. Furthermore, they integrated with Azure OpenAI for generating responses and constructed a FastAPI backend for querying the knowledge base.

Stemar's efforts resulted in a successfully implemented AI knowledge base, which enhanced the overall operational efficiency in the manufacturing environment. The project standouts include Stemar's source reference feature in responses and their Ragas-based evaluation pipeline for retrieval quality and answer reliability monitoring. Stemar had a commendable focus on addressing the nuances of industrial documentation, thus ensuring practical and reliable results. The use of the AI knowledge base has shown significant improvements in key quality metrics, such as Faithfulness and Recall.