GIAP

Inspiration

Managing procurement and inventory in small and medium businesses is a chaotic, manual process. In markets like Peru, this is compounded by strict, complex tax compliance regulations (such as SUNAT requirements). Traditional systems require rigid Excel imports, manual data entry, and constant cross-referencing with government tax portals to avoid penalties. We wanted to eliminate this busywork and create a system that adapts to the user and the local tax laws, allowing businesses to focus on growth instead of filling out forms and auditing suppliers.

What it does

GIAP is an intelligent purchasing agent powered by Google Gemini. It transforms reactive inventory management into a proactive, automated, and compliant process. Users interact with the system in natural language to create requirements. The AI automatically reads supplier Excel files and PDFs in any format, compares quotes, and predicts stockouts. Crucially, it acts as an automated compliance officer—cross-referencing supplier tax statuses in real-time (via SUNAT) and validating electronic invoices (XML) to ensure all procurement is legally sound and tax-deductible.

How we built it

We architected the system using a robust backend (Python/Node.js) and MySQL for data management, fully containerized with Docker for seamless deployment. The frontend is built with Next.js for a responsive UI. At the core of our logic, we integrated the Google Gemini API alongside local tax registry endpoints. Gemini processes natural language requests, parses unstructured supplier documents via its multimodal capabilities, and executes intelligent decision-making, while the backend ensures strict adherence to local invoicing rules.

Challenges we ran into

The biggest challenge was moving away from static database schemas to dynamic data ingestion while maintaining strict legal compliance. Parsing messy, unformatted Excel files and PDFs from different suppliers required heavy prompt engineering. We had to rely on Gemini's large context window to accurately map unstructured data into our relational MySQL database while ensuring tax variables (like IGV and retentions) were perfectly calculated.

Accomplishments that we're proud of

We successfully transformed a traditional, rigid module into a dynamic AI agent. We are extremely proud of the "zero-config and zero-risk" onboarding concept: users no longer need to manually map Excel columns, nor do they need to manually verify a supplier's legal status; the Gemini agent understands the semantic meaning of the data and audits the supplier instantly.

What we learned

We learned the immense power of multimodal LLMs in real-world business applications. Integrating Gemini taught us how to design prompt pipelines that act as dynamic ETL (Extract, Transform, Load) tools, bridging the gap between messy human documents and strict relational databases governed by complex local tax laws.

What's next for GIAP

Our next step is the 90-day sprint to launch. We plan to integrate WhatsApp for instant order approvals, implement the Gemini Vision API for direct photo-to-quote processing, and pilot the MVP with real industrial and commercial businesses to secure our first recurring revenue by solving both their procurement and tax compliance headaches.

Built With

• api
• docker
• google-apps-script
• google-gemini
• mysql
• next.js
• node.js
• python