I have always liked taking care of plants, but one problem I kept running into was that I would see them slowly wither, droop, or look unhealthy without really knowing why. Sometimes the leaves would turn yellow, sometimes the plant would look dry, and sometimes it just seemed “off,” but I could not tell whether the issue was temperature, humidity, watering habits, or a change in environment. I also tried plant apps that diagnose problems from pictures, but the advice was easy to forget soon after, and it did not give me a continuous sense of how my plant was doing over time.
That was the inspiration for AI Plant Companion. I wanted to build something that felt less like a one-time diagnosis tool and more like a small companion system that could continuously monitor a plant’s surroundings and translate sensor data into simple, understandable feedback. Instead of just giving raw numbers, I wanted the project to tell the user something meaningful, like whether the environment was comfortable or whether conditions might be too dry or too warm for a particular plant.
I built the project using an Arduino with environmental sensors to collect data such as temperature and humidity. On the software side, I designed the system so that the user can input the plant type, and the program interprets the sensor readings in a plant-specific way. The goal was to create a lightweight AI layer that converts measurements into natural-language plant status updates rather than just displaying sensor values.
Through this project, I learned a lot about how difficult it is to move from raw hardware data to something that actually feels useful to a person. I also learned that building a good user experience is just as important as getting the electronics to work. A sensor reading by itself is not very memorable, but a simple message that explains what it means is much more intuitive.
One of the biggest challenges was deciding how much I could realistically claim from the sensors I had. I originally wanted the system to capture more direct signals of plant health, but I had to rethink the scope and focus on environmental monitoring and interpretable feedback. Another challenge was making the project reliable enough for a demo while keeping the idea ambitious enough to be interesting. That process taught me how to balance creativity with feasibility, which was one of the most valuable parts of the experience.
In the end, this project became a way to combine something personal that I care about—keeping plants healthy—with sensing, embedded systems, and AI-driven interpretation. It is both a technical prototype and a response to a real problem I have faced myself: wanting to care for plants better, but not always knowing what they are trying to tell me.
Log in or sign up for Devpost to join the conversation.