Multilingual voice AI for government scheme eligibility
A voice-driven AI assistant bridging the information gap between India's farmers and 50+ government welfare schemes through natural voice interaction.
Sanchalak is an intelligent, multilingual, voice-driven assistant built to help Indian farmers instantly determine their eligibility for various government welfare schemes. The system offers a voice-first, regionally multilingual, and extremely low-latency (<5 seconds) conversational experience that bridges the digital divide. What makes Sanchalak uniquely powerful is its hybrid AI reasoning architecture, which combines a Prolog-based deterministic rule engine for 100% accurate and auditable eligibility decisions, a Large Language Model (Gemma) for natural, empathetic conversational responses, and seamless integration with speech-to-text (Whisper), machine translation (Azure), and text-to-speech (Azure) services. The system remembers a farmer's inputs across the session, supports highly diverse linguistic contexts, and is designed with real-world, rural deployment constraints in mind.
Streamlit, FastAPI, Python, SWI-Prolog, YAML, OpenAI Whisper, Azure Translator, Azure TTS, Gemma LLM, MongoDB, Docker, WebRTC