🚀 Inspiration In India, over 65% of the population lives in rural areas where digital literacy is limited, access to government schemes is complex, and online security is often overlooked. We envisioned SaarthiNet as a way to democratize access to digital infrastructure—empowering every Indian, regardless of literacy or location, to interact with technology using their voice in their native language, while staying protected from rising cyber threats.
💡 What it does SaarthiNet is a voice-first AI assistant combined with a real-time cybersecurity mesh that enables:
Voice-based access to government schemes, banking, health services, and verified information in regional languages.
RakshaNet, our AI-powered cybersecurity shield, which detects phishing, scams, and fraud in real-time.
Offline-to-online gateway, allowing users to interact with digital services through voice without needing high-end smartphones or constant internet.
Think of it as “Aadhaar + Google Assistant + Cybersecurity, made for Bharat.”
🛠️ How we built it Voice AI & NLP: We trained multilingual models using Hugging Face Transformers and integrated open-source ASR (speech-to-text) with custom NLP intents for local dialects.
Cybersecurity Mesh: Built using Python, Snort, and custom rule-based + ML models for real-time fraud/phishing detection.
Frontend: A lightweight ReactJS PWA for admins and rural digital centers.
Backend: Node.js + MongoDB for the core logic, with Firebase and Twilio APIs for SMS/IVR-based fallback access.
Deployment: Dockerized services on AWS EC2 with Kubernetes for scalability.
🧱 Challenges we ran into Dialect & Accent Variability: Indian languages are complex, and voice recognition struggled with regional accents. We mitigated this using custom data and fine-tuning.
Low Connectivity: Designing a system that works in low-bandwidth areas required fallback IVR/SMS support and local caching.
Security vs Usability: Integrating strong security features without making the experience complex for users unfamiliar with digital tools.
🏆 Accomplishments that we're proud of Built a functional prototype that supports voice queries in Hindi, Bhojpuri, and Marathi.
Real-time scam/phishing detection demo that blocked malicious links on the fly.
Created a working IVR + voice bot that fetched real-time PM-Kisan scheme eligibility for a mock Aadhaar ID.
Positive feedback from rural community testers on usability and relevance.
📚 What we learned Empathy-first design is crucial when building for Bharat—what works in cities doesn’t always translate to villages.
Open-source voice tech can be powerful, but local data collection and fine-tuning are key.
Bridging access and security is not just a tech challenge—it’s a behavioral and cultural one too.
🔮 What's next for SaarthiNet Expand voice support to 10+ Indian languages and dialects with local voice data contributions.
Partnerships with Digital Seva Kendras and NGOs to pilot in real rural districts.
Launch RakshaNet API for fintech apps, ed-tech platforms, and digital banks.
Apply for government & CSR grants to scale voice + cybersecurity awareness in rural India.
Build SaarthiNet devices – low-cost speaker-based systems (like Alexa) powered by local servers for offline usage.
Log in or sign up for Devpost to join the conversation.