What started as an idea in my head has just crossed 10,000+ users.
Introducing AgroSage.
No course inspired this. No tutorial to copy from. I identified a real problem, looked at what already existed to solve it, decided none of it fully fit , and decided to BUILD MY OWN.
The idea came together over 2 - 3 months of thinking ( Rather, I would say Overthinking : D ) through the problem and the architecture . I started actually building it on July 13, 2025.
After completion . I posted it across a few platforms on the internet , no marketing plan even nothing was organized . I genuinely did not expect it to cross even 1,000 views. Somehow, it kept growing and boom , Today : 10,000+ unique users.
The problem AgroSage solves : Most farmers still decide what to plant based on tradition or generic advice that is not localized to their actual soil and climate over their area . The data that could make this decision better i.e. soil pH, annual rainfall, temperature, relative humidity , exists in scientific databases, but it is not in a form anyone without a GIS background can turn into a meaningful decision for planting.
How it works - Simple two layers structure :
Machine Learning : A classification model trained on 2,200+ agricultural records predicts the best suited crop ( from 30+ categories ) for any location. It is backed by 91.8% test accuracy on a completely unseen dataset of 100+ records , marking a strong check on how the model actually generalizes . It runs on live soil and weather data pulled automatically from the SoilGrids API and NASA POWER APIs. How it works : Simple Choose a date and either drop a pin on a map or search a location or enter coordinates manually or just enter your local conditions directly.
Generative AI : Once weather parameters are fetched and a crop is selected , a LangChain + Groq advisory layer converts that single prediction into 5 specific cultivation precautions , tailored to the exact soil pH, rainfall, temperature, relative humidity and sowing date for that spot, aimed at maximizing yield.
What this integration taught me : - ML decides what to grow. AI helps you grow it.
This is just a milestone that means a lot to me. A project that started as a CV line has turned into something people actually use. If you try it and think something is missing or something that could make it more interesting . I would genuinely love to hear your ideas on what to add next.
Try it: agrosage-ai.streamlit.app
#MachineLearning #ArtificialIntelligence #AgriTech #PrecisionAgriculture #SmartAgriculture #DataScience #DataDrivenAgriculture #GenerativeAI #ClimateTech #ClimateSmartAgriculture #EnvironmentalEngineering #AIforGood #TechnologyForFarmers #StudentInnovation #BuildInPublic
