Propose Architecture for Marathi language based IPA ( Voice assistance app )

I want to create farmers app which will be used by many farmers in Maharashtra. They speak Marathi. App is expected to be used by 1 millions farmers in Maharashtra. I will be kind of voice assistance chatbot which will ask couple of quesitons to the farmers on daily basic about farming. Like what he is going to do today, what seeds he is going plant, what fertilizer he is going to use, how much amount of seeds he is going to plant. Basically his daily details about farming schedule. So that we get idea what farmer is doing and what will be the expected date of getting crops ready. Also we get idea about quality of crop from type of seeds and fertilizer he has used. So that will give us idea about when crop will be ready and how much production we will have. That way we can connect the production with customer to feel suppy chain gap. I want to create robust, scalable, secure architecture for this application. I already have idea about tech stack required at good extenese. But getting someone having extensive experience in building architecture will be helpful. Rough technology flow is here - 1. The user interacts with the app through their mobile device, using the Marathi language for voice input. 2. The voice input is sent to a speech-to-text service, such as Google's Speech-to-Text API, which converts the speech to text in Marathi. 3. The text is then sent to a natural language processing (NLP) service, such as Google Cloud Natural Language API, which is used to extract relevant information such as seed and fertilizer names. 4. The extracted data is then sent to a NoSQL database, such as MongoDB or Cassandra, for storage. 5. The backend is built on a serverless architecture, such as AWS Lambda or Google Cloud Functions. These functions are triggered by the user's interactions with the app and handle tasks such as speech-to-text conversion, NLP, and data storage. 6. The entire system is hosted on a cloud platform like AWS or Google Cloud, which allows for scalability and cost-effectiveness. Can you propose an architecture for this kind of app. This work can be long-term. And if you are good I can keep taking your help on regular basis.