

Proud to share my capstone project from the 5-Day AI Agents Intensive with Google, hosted on Kaggle, one of the most competitive global platforms for applied AI.
This wasn’t just a course — it was a worldwide hackathon focused on building agent-based systems that solve real human problems.
🌾 The Problem
Farming decisions today are still driven by guesswork, fragmented advice, and outdated information.
Farmers don’t think in dashboards or dropdowns.
They think in questions:
“What crop should I grow next season?”
“Should I irrigate today if it might rain tomorrow?”
“Is my soil suitable for vegetables?”
Most existing agri-apps fail because they force farmers to adapt to software, instead of adapting software to how humans naturally think and speak.
🤖 The Solution — KRISHI-MITRA
KRISHI-MITRA is an AI-agent–based assistant designed to act as a digital companion for farmers, not just another app.
Built using Google Gemini + Agent Development Kit (ADK), the system goes beyond chatbots.
It can:
1) Understand free-form natural language
2) Remember farmer and farm context persistently
3) Register farmers and farms conversationally
4) Recommend crops based on soil, history, and trends
5) Create reminders and alerts
6) Maintain memory across sessions using a database
7) Reason about multi-factor agricultural decisions
No forms.
No rigid workflows.
Just conversation → reasoning → action.
🧠 Why Agents (not traditional apps)
Traditional UIs break when workflows become human and continuous.
Agents excel because they:
hold memory,
adapt over time,
call tools autonomously,
feel natural even to non-technical users.
KRISHI-MITRA demonstrates how agentic AI can turn complex, real-world systems into simple conversations.
🛠️ Tech Highlights -
Google Gemini 2.5 Flash Lite ,
ADK-based agent orchestration ,
Custom tool calling with schemas ,
Persistent memory via SQLite (production-ready upgrade path) ,
Debuggable, inspectable agent execution ,
End-to-end conversational demo.
🌍 Why this matters
This isn’t just an academic project.
It’s a proof-of-concept for how AI agents can empower farmers, improve decision-making, and make advanced technology accessible to communities that need it most.
This project made one thing very clear to me:
Agents are not the future of apps — they are the replacement for them in many real-world domains.
Grateful to Kaggle and Google for the platform.
More importantly, excited about building systems that solve real problems, for real people.
🎥 Demo video attached below.
#AIAgents #AgenticAI #Google #Kaggle #Hackathon #AIForGood #AppliedAI #LLMs #StartupIdeas #AgriTech #SystemDesign #Builders
19 Janvier 2026 à 15h16
Voir sur LinkedIn →