Making scientific data more accessible.
Problem
Excelra, a global leader in life sciences informatics, faced a challenge with their GOSTAR database. Medicinal chemists and researchers were struggling with traditional search methods that were complex and technical.
Despite having valuable data that could accelerate drug discovery, users found it difficult to extract the insights they needed. They needed a way to make their vast scientific datasets more accessible without sacrificing technical depth.
Solution
We developed an AI-driven Natural Language CoPilot that transforms how researchers interact with the GOSTAR platform. Built with a modular interface and serverless backend, the system lets users query complex scientific data using natural language.
We integrated Elasticsearch and ClickHouse for rapid data retrieval, while supporting complex queries and SQL generation. The platform includes user feedback mechanisms and flexible output formats, making it easier for researchers to work with their findings.
Results
The CoPilot changed how researchers interact with scientific data. Complex queries that once required technical expertise became conversational. The system proved robust in user acceptance testing, handling diverse and complex scientific queries while maintaining accuracy. Most importantly, researchers could now focus on generating insights rather than wrestling with query syntax.
Technologies
