HUGO Agent - A Memorization LLM

Hugo connects ERP, emails, and contracts to flag risks, suggest actions, and deliver dashboards and alerts—keeping Voltway’s procurement team focused and in control.

Comment

Inspiration

We were inspired by the Dryft company challenge and the idea of creating a smart, adaptable, and straightforward assistant that could ease customer service and operational work.
Our goal was to build Hugo — an AI-powered co-pilot that would pull data from multiple sources like ERP systems, supplier emails, and contracts, organize it intelligently, and allow company teams to focus more on solving problems rather than chasing information.

What it does

  • AI-driven analysis: Provides insights into procurement processes and identifies potential areas for improvement.
  • Data summarization: Organizes large volumes of data into concise, actionable summaries.

How we built it

  • ChatGPT: We used the OpenAI ChatGPT tool to generate realistic and simple UI design mockup for Hugo, showcasing how the assistant would look and function.
  • Google Colab: This platform was utilized to demonstrate a basic implementation of Hugo’s functionality and to share our working code in an interactive environment.
  • OpenAI: Provider of the significant LLM APIs used in the demonstration. These APIs were critical in enabling Hugo’s AI capabilities, like natural language processing and smart summarization.
  • Dryft: Dryft provided the original idea and the basic starter data files, giving us a foundation to begin developing Hugo’s capabilities.
  • Anvil: Used to demo our project as a web app.

Challenges we ran into

  • Task Integration into AI Workflow: We needed to carefully map procurement tasks to Hugo’s AI capabilities.
  • Design Approach: We wanted to efficiently integrate various sources of data (emails, ERP systems) without compromising accuracy. We had to carefully design a solution that pulled relevant information from each data source and processed it in a way that remained clear.

What we learned

  • API Handling: How to design Hugo to effectively suggest next steps while remaining concise and on-point. We created responsive integration between the data and AI output, while minimizing conflicts or redundancy.

Built With

Share this project:

Updates