Anganwadi Schedule Optimiser.

Project Story: Anganwadi Schedule Optimiser 💡 Inspiration India’s 13.9 lakh Anganwadi centers are the backbone of rural healthcare, but they are drowning in administrative chaos. During our research, we found that missed immunizations are rarely due to lack of supply—they are due to scheduling conflicts.

Workers often find out about a village medical camp or an ASHA worker visit only minutes before it happens, forcing them to cancel preschool nutrition or education sessions. With low keyboard literacy and high workloads, manual registers simply can’t keep up. We wanted to build a tool that feels as natural as a conversation but has the logic of a master logistics engine.

🚀 How we built it We prioritized a "Voice-In, Schedule-Out" architecture using the Gemini 3 Flash model.

Multimodal Parsing: We used Gemini’s native audio understanding to skip traditional Speech-to-Text. The model directly "hears" the intent in Marathi or Hindi.

Agentic Decision Making: We implemented Function Calling to extract structured JSON data from messy, natural-language voice notes.

The Constraint Solver: We built a custom backend logic where tasks are categorized as Fixed (e.g., Vaccine Camps) or Flexible (e.g., Weight Checks). If a Fixed event is added to a busy slot, the system automatically re-calculates the week and "bumps" flexible tasks to the next available free slot.

Premium PWA: The frontend was built with React and Tailwind CSS, featuring a high-contrast dark theme designed specifically for visibility on low-end Android devices in bright field conditions.

🧠 Challenges we faced The biggest hurdle was multilingual conflict resolution. A worker might say, "Polio camp mangalwari aahe" (Polio camp is on Tuesday), and the AI needs to know that "Tuesday" is already occupied by a nutrition check. We spent significant time tuning the system instructions to ensure the "Constraint Solver" always prioritized life-saving medical interventions without losing the daily nutrition tasks. We also navigated Firestore indexing hurdles to ensure real-time performance across 2G village networks.

📈 What we learned We learned that AI’s true power in social impact isn't just "chatting"—it's structured reasoning. By offloading the "thinking" part of scheduling to Gemini, we can reduce a worker's administrative burden by an estimated 70%, allowing them to focus on what matters most: the children.

🏆 What's next for Anganwadi Schedule Optimiser Our goal is to integrate this directly into WhatsApp and USSD, allowing workers to send voice notes without needing a dedicated app. We also plan to add Travel Logistics, where Gemini blocks out morning or afternoon slots based on the "hamlet distance" mentioned in the voice note.

Built With

• cloud-firestore
• firebase-authentication
• gemini-3-flash
• gemini-function-calling
• lucide-react
• progressive-web-apps-(pwa)
• react.js
• tailwind-css
• vite