AgroLatam Agent

Inspiration:

The inspiration for Agente AgroLatam stems from the reality of the 60 million farmers in Latin America who remain disconnected from digital market intelligence. Growing up in Peru, I witnessed firsthand how small-scale producers often lack access to the real-time data needed to compete in global markets. I wanted to bridge this gap by building an autonomous system that acts not just as a tool, but as a proactive, intelligent partner for agricultural decision-making.

What it does:

AgroLatam is an autonomous AI ecosystem that democratizes access to strategic data for 18 countries. It functions as a "pocket agronomist," monitoring commodity markets, climate data, and export opportunities in real-time. The agent proactively alerts farmers to price volatility, warns of climate risks affecting harvest windows, identifies international export matches, and automates the creation of complex agricultural credit documentation.

How we built it:

The project was built using a robust modern stack:

Intelligence: Gemini 2.0 Flash powers the agent's reasoning capabilities, integrated via Google Cloud Agent Builder.

Data Integration: Fivetran serves as the backbone for our data pipelines, ensuring clean and reliable market data ingestion.

Backend: FastAPI handles high-concurrency requests and asychronous processing.

Deployment: The application is deployed on Cloud Run and Vercel for scalable, low-latency performance.

Challenges we ran into:

The primary hurdle was managing high-latency data ingestion from multiple, disparate sources across 18 different countries. Ensuring that the agent could synthesize real-time market data with localized climate indices without slowing down the user experience required significant optimization of our asynchronous backend logic and API handling.

Accomplishments that we're proud of: I am particularly proud of the successful integration of Fivetran for seamless data synchronization, which transformed a fragmented set of data points into a cohesive, actionable dashboard. Furthermore, achieving a functional "Autonomous Agent" flow that can autonomously generate credit documentation represents a major step forward in reducing the bureaucratic burden for farmers.

What we learned:

Building this project deepened my expertise in Autonomous Agent Orchestration. I learned how to bridge the gap between technical infrastructure—like real-time data pipelines—and the human-centric needs of the agricultural sector. This project also significantly improved my proficiency in architecting scalable AI-driven backends using FastAPI and integrating them into the Google Cloud ecosystem.

What's next for AgroLatam Agent:

The next phase focuses on scaling the agent to include more niche crop data and integrating satellite imagery for advanced crop health monitoring. I plan to expand the agent's capabilities to include a peer-to-peer marketplace module, allowing farmers to coordinate logistics directly within the platform, further reducing reliance on intermediaries.

Tip para Devpost:

Si tienes capturas de pantalla de tu dashboard funcionando, asegúrate de subirlas junto a este texto en la sección de "Gallery" de tu proyecto. ¡Esto ayudará muchísimo a que los jueces visualicen el impacto de tu trabajo!