warehouse

Case

Optilogic

Retail

An AI chat agent for designing supply chains

GenAI / agentsNLPProcess efficiencyUser experience

Modulai built Leapfrog AI, a custom LLM agent embedded in Optilogic's Cosmic Frog platform for supply chain design. It lets users work with supply chain data in plain language, querying data, running SQL operations, and managing scenarios through conversation rather than code.


  • Challenge

    Working with Cosmic Frog meant writing complex SQL queries to get insights out of the platform. That put a technical barrier between business users and their own data, slowing decisions and leaving non-technical teams dependent on IT. Optilogic wanted a way for those users to interact with supply chain data directly, without needing to write queries themselves.

  • Solution

    Leapfrog AI is embedded within Cosmic Frog to transform natural language queries into precise SQL commands and actionable insights. It supports tasks such as querying data, running scenarios, geocoding, and exporting results all while ensuring safe execution through robust validation using Pydantic and comprehensive error handling. The conversational interface bridges the gap between complex data analysis and everyday business language.

  • Tools

    Leapfrog AI leverages OpenAI's GPT-models for natural language understanding and SQL generation, Pydantic for schema and function validation, and PostgreSQL for data execution. It also incorporates multilingual NLP processing, automated data diversification for robust training, and real-time monitoring to ensure reliable, secure operations across global supply chain applications.

  • Value created

    Leapfrog AI is integrated into Optilogic's Cosmic Frog platform, giving its users a way to work with supply chain data directly instead of routing requests through technical staff. By turning everyday questions into validated SQL, it lets business users run their own queries and scenarios, shortening the path from question to answer in workflows that previously required someone who could write the query.