FarmFlight

Inspiration

The agriculture industry has morphed into a swamp of centralization, where industrial farmers, who make up only 15% of the farming industry, control the vast majority of the market (85%). Smaller family farmers often lack the liquidity to keep up with expensive, new technological developments.

These farmers are the backbone of the world’s working class, providing us with fresh, nutritious food while taking home very little for themselves. Furthermore, farming practices as a whole are extremely inefficient (e.g. up to 40% of the water used for agriculture is wasted due to inefficient irrigation practices and improper water management).

What if we could provide a helping hand to family farmers while improving resource efficiencies?

Meet FarmFlight, the all-in-one drone and software system that empowers family farmers with real-time agricultural data and aerial crop management.

What it does

FarmFlight is a Gemini-powered dashboard and drone system that puts the power back in the hands of family farmers. It features a Gemini chatbot incorporating numerous data sources, weather alert system, and guided actions that can be carried out via the paired drone. The drone is capable of user-controlled flight, surveillance, and object dispersal. This allows farmers to collect footage of their fields with its mounted camera, as well as disperse seeds and fertilizers with ease. The included Gemini chatbot uses Retrieval-Augmented Generation from two online communities of farmers with a combined monthly membership of over 100,000 people to provide the most insightful and applicable responses. The chatbot also references user-uploaded drone footage of their farms, giving specific insights into their crop health.

How we built it

FarmFlight is built around the core functionalities of providing live updates about a farmer’s crops and providing the hardware necessary to augment the agricultural workflow. The dashboard frontend is constructed using Next.js with TypeScript and TailwindCSS for modular, modern design. We queried National Weather Service APIs to deliver live weather updates and advisories. Our chatbot is powered by Google Gemini and is enhanced by agricultural forum posts gathered through web scraping using Octoparse. The scraped data is then stored as 384-dimension vector embeddings using the Qwen GTE-small to the PostgreSQL database, hosted on Supabase. Video upload is handled by S3 buckets. This allows the videos to be stored and viewed, as well as analyzed using Gemini’s vision-to-text capabilities. The dashboard is deployed using Vercel.

Our hardware goal was to create a drone capable of crop dusting and data collection, to complement the FarmFlight dashboard. The process was mainly concerned with reducing the torque and weight on the drone. Given that we were modifying a cheaper consumer drone, we had extremely stringent weight constraints (about 70 grams in addition to the drone itself). We utilized the drone’s built-in camera servo, attaching a dispenser system consisting of a 3d printed container and rubber band to the camera and frame of the drone.

Challenges we ran into

While testing and modifying the consumer drone we chose to utilize, we found that it struggled to support heavy loads like the servo-based seed distributor we originally planned to use. After spending eight hours trying to make an electronic dusting system using an Arduino, servo, and 9V battery, we realized that we could not afford to bring on our own battery and electronics due to their additional weight. This led us to 3D print a custom lightweight tray to support and distribute the seeds while maintaining the drone’s airworthiness. Another challenge we faced was how to retrieve and aggregate environmental data to inform our chatbot’s recommendations. We initially wanted to use satellite imagery to analyze the soil and other factors, but found that these images were too low quality to provide meaningful information. This led us to use AI-powered web scraping on agricultural forums to provide an alternative data source. With thousands of entries at its disposal, Gemini is able to dispense meaningful advice to users.

Accomplishments that we're proud of

Seeing as this is our team’s first hackathon together, we are extremely proud of how efficiently we worked together. Three of our members were in their first hackathon. Our team was able to divide and conquer with success, creating a solution that blends hardware and software into a suite of tools for farmers. We are also proud of how many different technologies we were able to integrate into our project, from CAD, to web scraping, to RAG.

What we learned

As a team, we learned how to pivot rapidly when a decision was not panning out. This flexibility allowed us to maintain momentum throughout the hackathon. We also learned the importance of user-centered design, which should be simple, consolidated, and easy to use. This is especially relevant when creating a product that is meant to ease the user’s quality of life or work. We also learned how to combine different skillsets, be it hardware, software, communication, or NLP.

What's next for FarmFlight

In the near future, we would love to meet with farmers and stakeholders to determine how we can better meet their needs, as well as expand the impact of our project. We hope that FarmFlight can be distributed to family farmers across the country to help them increase efficiency and promote sustainability throughout the farming industry.