Toptal revolutionizes brewing company’s document processing with 98.7% accuracy.
Facing inefficiencies in extracting knowledge from unstructured documents, a leading brewing company enlisted Toptal to enhance a business manager-focused chatbot.
Client
A global brewing company producing beers and beverages with a portfolio of iconic brands worldwide.
Employees
Revenue
Industry
Consumer Packaged GoodsDelivered Services
Get a free consultation now
Schedule a CallChallenge
Developing a chatbot exposed the need for an LLM specialist for complex RAG tasks, with an external API raising costs and causing latency delays.
Solution
Advanced Document Processing
Toptal crafted a model to effectively identify tables, charts, and figures in images, streamlining document parsing with support from a PDF reader.
Edge-optimized Performance
Toptal refined a YOLO (you only look once) model specific to the company’s requirements, resulting in a 10x speed increase and fivefold accuracy improvement.
Outcome
Enhanced Data Retrieval
The new table detection model increased its recall rate to 98.7%, improving data retrieval accuracy and providing managers faster access to vital information.
Boosted Efficiency and Savings
The edge-based NER model decreased latency by 3.5 seconds per request, significantly improving operational efficiency and reducing token usage costs.
Get a free consultation now
Schedule a Call