Developing a voice-first super application for a US-based AI technology company

The key objective of the Daffodil team was to develop sub-apps that would enhance the end-user experience in all ways possible. The entire code repository was audited to understand the factors influencing the performance of the super app. Sufficient R&D was carried out to come up with the most interactive UI solutions which were up to the latest global UX standards. The sub-apps were built over the Flutter UI framework with the Dart programming language for platform-agnosticity. Flutter was also chosen for the creation of reusable widgets for common use among all sub-apps to ensure consistency across the super app.

Developing the Restaurant Sub-app

Our team built the restaurant sub-app from the ground up by creating a mechanism for aggregating restaurant-specific data using Python-based web scraping. This helped automate the repetitive parts of searching for a restaurant online, speeding up the data collection process by writing the code for it only once. Node APIs were used to serve the restaurant data to the frontend on Flutter. The following capabilities were built into this sub-app:

• It suggests restaurants in the order of proximity to the user’s location
• It lets users mark restaurants as favorites and get future recommendations accordingly.
• The AI voice assistant asks questions like where the user wants to go and what cuisine they prefer to bring up restaurant listings.
• The AI also lets users make reservations directly from the sub-app.
• Over several interactions with the user, the AI learns the user’s preferences with respect to location, cuisine, and meal timings to make better recommendations.

Transforming the Weather Sub-App

This existing sub-app was analyzed using heatmaps and our team found that its UI was cluttered, unappealing and unresponsive. The sub-app was not easy to navigate and had performance issues. The team resolved these issues in the following ways:

• Several fixes were rolled out for rendering issues and the design was stabilized. The UI was sharpened and brought up to UX industry standards. Better animations and motion to UI elements were added to enhance the interactivity of the sub-app.
• An AI voice assistant was built into this sub-app, allowing users to ask a series of questions about the weather and get instant results.
• The Weather sub-app was linked to the Restaurant sub-app, providing users with live updates about the weather around the restaurant that they are planning to visit.
• Additionally, the team found and resolved about 90% of security vulnerabilities in the Weather sub-app.

The AI Voice Assistant

AI-enabled Automated Speech Recognition (ASR) was integrated into both the sub-apps for retrieving weather and restaurant-related information. Our developers made the AI voice assistant highly conversational in nature with a rich set of voice queries. The wide variety of voice commands included around 50 voice queries with over 250 variations. The frontend was connected to Google speech2text modules. Machine Learning (ML) based intention detection and Natural Language Processing ensured that accents, emphasis, and inflections in the user’s speech could be understood by the AI. 

The perceived intelligence of the AI could increase after every interaction with a user and the relevance of information will increase each time. The AI will ask appropriate questions like where the user would like to go, what type of cuisine they’d like, and help make reservations directly at the restaurant of their choice, while at the same time learning the user’s preferences for making recommendations in the future.

Database And Reusable Widgets

All the app processes take place on a lightweight and fast database, HiveDB, ensuring a short response time in accumulating restaurant and weather search results. Daffodil built custom algorithms to handle mapping issues between the varied Restaurant APIs. Several reusable UI widgets were also created in Flutter for common use among all the sub-apps, in addition to the Restaurant and Weather ones. This would ensure consistency throughout the super app.