FarmAID

• 

  Final Result

Inspiration

We were inspired by the challenges farmers face in regions with poor internet connectivity. Many lack immediate access to expert advice on crops, irrigation, and pest management, which directly affects their yield and income. We wanted to create a local AI assistant that provides reliable guidance in multiple languages, making smart farming accessible to everyone, even offline.

What it does

FarmAid is an offline, multilingual AI assistant for farmers:

• Provides advice on crop care, irrigation, and pest management
  
• Works entirely offline, running locally on LM Studio or Ollama
  
• Supports English and Urdu (with plans for more languages)
  
• Offers reasoned, actionable guidance, not just generic answers
  

How we built it

• Used GPT-OSS-20B, an open-source large language model
  
• Integrated with LM Studio / Ollama for lightweight offline deployment
  
• Implemented multilingual support for English and Urdu
  
• Designed a farmer-friendly interface optimized for low-resource devices
  
• We experimented with formulas for fertilizer optimization, like:

$$ F = \frac{N \times A}{Y} $$

Where:

• ( F ) = Fertilizer needed (kg)
  
• ( N ) = Nutrient requirement per hectare
  
• ( A ) = Area of the field (hectares)
  
• ( Y ) = Expected yield (tons)
  

Challenges we ran into

• Ensuring the model runs smoothly offline without large hardware requirements
  
• Translating complex agronomy knowledge into simple, actionable language
  
• Optimizing performance vs. size to make the model usable on average laptops
  

Accomplishments that we're proud of

• Successfully deployed an offline AI agent capable of giving intelligent farming advice
  
• Achieved multilingual support in English and Urdu
  
• Built a domain-specific AI that can help farmers increase yield and save resources
  

What we learned

• Offline AI deployment is feasible for real-world applications
  
• Farmers need advice that is clear, actionable, and context-aware
  
• Multilingual support is essential for accessibility and adoption
  
• Working with domain-specific AI requires careful knowledge curation
  

What's next for FarmAid

• Add support for more local languages like Punjabi, Sindhi, and Pashto and "international languages*
• Integrate image recognition for crop disease detection
  
• Build a mobile-friendly offline interface for field usage
  
• Partner with agri-tech companies for testing and scaling

Built With

• gpt-oss-20b
• lmstudio
• ollama
• ollama-**apis-/-models:**-gpt-oss-20b-(open-source-large-language-model)-**databases:**-local-json/sqlite-for-caching-queries-(offline)-**other-tools:**-ffmpeg-(for-demo-video)
• openai-compatible
• python
• qwen3-4b
• restapi