Inspiration
The inspiration behind Agrolytix came from witnessing the struggles of Indian farmers — from misinformed crop selection to language barriers in accessing expert advice. I was determined to build an AI system that could understand their needs, speak their language, and provide real-time, intelligent support. The vision: empower every farmer with a digital assistant that’s local, smart, and always accessible.
What it does
Agrolytix is a multilingual, AI-powered farming assistant that helps farmers with:
Soil analysis based on pH, moisture, and fertility
Crop recommendations tailored to local conditions
Voice interaction in Marathi, Hindi, and English
Real-time chatbot that answers farming questions
Region-wise crop planning based on weather and soil
It combines the power of a fine-tuned AI language model with a user-friendly voice and chat interface, all tailored to Indian agriculture.
How we built it
Agrolytix was developed using a combination of open-source AI models, agricultural data, and real-world farmer input:
Collected Q&A datasets from ICAR, Krishi Vigyan Kendra, and YouTube (transcribed to text)
Fine-tuned LLMs like Mistral, Phi-3, and LLaMA 3 with HuggingFace + LoRA
Used Whisper AI for voice-to-text and Coqui TTS for text-to-speech in local languages
Built the frontend with Gradio, Streamlit, and Flutter
Deployed using LM Studio, Ollama, and HuggingFace Spaces
Challenges we ran into Maintaining multilingual accuracy across Marathi, Hindi, and English
Building a clean and diverse agricultural dataset for fine-tuning
Designing a low-bandwidth, offline-ready interface
Training the AI to understand hyper-local queries with real context
Accomplishments that we're proud of Developed a live working prototype for soil-to-crop AI assistance
Enabled voice-based farmer interaction with regional language fluency
Launched Agrobot & Technologies as a full-fledged AI agri-tech startup
Honored by NITI Aayog, selected as Meta Student Ambassador, and recognized as one of the youngest innovators in AI
What we learned
How to fine-tune LLMs with local datasets
Building AI that respects language, culture, and usability
Leveraging open tools like HuggingFace, Coqui, Whisper for smart farming
That solving real-world problems is about making tech simple and human
What's next for Agrolytix by Agrobot & Technologies
Releasing the Agrolytix app with multilingual voice & chat
Expanding datasets for state-wise farming recommendations
Integrating IoT soil sensors and automating diagnosis
Partnering with government and agri-bodies to scale impact
Launching offline-ready versions for rural and low-connectivity zones
Making Agrolytix the voice of Indian agriculture
Built With
- android-studio
- colab
- css
- csv
- dart
- figma
- firebase
- flask
- flutter
- github
- googlesheets
- gradio
- html
- huggingface
- javascript
- json
- kaggle
- llama
- lmstudio
- mistral
- mit-app-inventor
- ollama
- paperspace
- phi-3
- python
- speechapi
- streamlit
- transformers
- translateapi
- webflow
- whisper
- wordpress
Log in or sign up for Devpost to join the conversation.