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

100,000+

Revenue

$55.2B

Industry

Consumer Packaged Goods

Delivered Services

Get a free consultation now

Schedule a Call

Challenge

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.

Share

Get a free consultation now

Schedule a Call