KisanMitr
Your Digital Companion in Every Harvest.
Created on 8th February 2026
•
KisanMitr
Your Digital Companion in Every Harvest.
The problem KisanMitr solves
Farmers struggle with accessing timely, reliable information for day-to-day decisions related to crops, weather, pricing, and government schemes. Most digital tools assume internet access, smartphone literacy, and English proficiency, which many farmers do not have. This gap leads to poor decision-making, financial losses, and dependency on middlemen.
What People Can Use It For
- Voice-First Assistance
Farmers can speak in their preferred language (Hindi / English / Punjabi) to ask questions about crops, weather, inputs, or profits without navigating complex apps.
- Weather & Location-Based Guidance
Provides real-time, location-specific insights to help farmers plan sowing, irrigation, and harvesting more safely and accurately.
- Inventory & Orders Awareness
Farmers can easily track listed produce and received orders using voice or simple inputs, removing the need for manual record-keeping.
- Schemes, Loans & Policy Clarity
Explains government schemes, loans, and agricultural laws in simple language and shows how they directly impact the farmer.
- Continuous, Context-Aware Support
Allows follow-up questions and personalized guidance based on season, location, and farmer needs.
Challenges we ran into
1. Integrating a Voice AI Agent with Calls
One major challenge was building a reliable voice-based AI agent that could work over phone calls. Connecting Twilio, call flows, and AI responses required strict formatting, valid webhooks, and deployed endpoints.
How I overcame it: I broke the problem into smaller steps—first testing call triggers, then static responses, and finally AI-driven replies. This helped isolate failures quickly and understand where the integration was breaking.
2. API & Quota Limitations
While experimenting with different AI APIs (Gemini, Codex, etc.), I frequently hit quota limits, unsupported models, and invalid API key errors. This slowed down iteration and testing.
How I overcame it: I designed graceful fallbacks, used cached responses where possible, and explored free/open-source alternatives to reduce dependency on paid APIs.
3. Multilingual Support Across the App
Making the entire system language-aware (UI, chatbot, and voice agent) was harder than expected, especially ensuring consistent translations and correct intent detection across languages.
How I overcame it: I centralized language selection early in the flow and passed the chosen language as context to every AI and API call instead of handling it separately in each feature.
4. Merge Conflicts & Rapid Iteration
Frequent feature additions led to Git merge conflicts, especially in shared files like translations and dashboards.
How I overcame it: I resolved conflicts manually by understanding intent rather than auto-merging, and started committing smaller, isolated changes to reduce overlap.
Tracks Applied (3)
Google Gemini
Major League Hacking
ElevenLabs
Major League Hacking
MongoDB Atlas
Major League Hacking