Digital product · Proposal with prototype
IAgricultura
Academic pair project that proposes an assistant for small producers, with an experience built on crop, climate and soil information. The stage presented includes the design of the solution and an interface prototype with eight screens; no model has been trained and there is no operational integration yet.
- Type
- Academic pair project · Data Science / FIAP
- Status
- Proposal and interface prototype
- Scope
- Assistant for small producers
The problem
Many small farmers grow what their family has always grown. When the climate shifts, they lack information about other crops their land and terrain could support, and about how to manage, process and sell them.
The project cites a figure from Jornal da USP: family farming accounts for 77% of producers and 23% of agricultural income.
The proposal
The farmer describes their situation by voice. The system transcribes it, cross-references climate, location and time of year, and suggests suitable crops along with video lessons on planting and management. The interface was designed for people with little schooling, leaning heavily on images, video and voice.
- Farmer audio
- Transcription
- Climate, soil and season
- Recommendation model
- Suggested crop
- Management video lessons
App prototype
Eight screens designed for the farmer: login, menu, a voice and video chat with the assistant, diagnostics, suggested videos, crop tracking and the property data. The screens are an interface prototype: browse them with the tabs; the buttons drawn on them do not work as an app.
Proposed architecture
The design separates what is processed in batch from what needs a fast response, in four layers.
Data capture
IBGE (municipal agricultural production), Embrapa’s Pronasolos and data generated by the farmers themselves.
Processing
A batch layer with Excel, Power BI and Python; a speed layer for quick AI answers and database queries.
Serving layer
Database, tables, AI models and an API.
Consumption
A Power BI dashboard, the app, the voice assistant and the personalised recommendation.
Planned technologies and sources
- Batch layerplanned
- Serving layerplanned
- Data sourcesplanned