HeyBuddy mobile app development

An AI Companion chat app built for iOS on a scalable Serverless backend

See more

The Brief

Build a strong foundation of an iOS chat app for AI companions that stands out among its competitors in price efficiency and scalability.

The Brief

Build a strong foundation of an iOS chat app for AI companions that stands out among its competitors in price efficiency and scalability.

Sound unit economics

LLM API costs are significant and quickly add up. This is especially true in the context of an AI chat interface with frequent interactions. Our goal was to establish a backend infrastructure that remains cost-efficient irrespective of usage volume. At the same time we aimed to optimise LLM prompting, while balancing the contextual information an LLM needs for genuine dialogues.

Sound unit economics

AI that feels human

Often chatbots have an "at your service" attitude. This is great for utility workers, but conversations aren't one-sided, and require a kind of push and pull. Both conversation partners should be able to initiate the conversation and introduce topics freely. Our goal was to create this calibre of AI. One that has a point of view and can talk to you at any time they wish.

AI that feels human

The Serverless Architecture

We built a native iOS app written in Swift, connected to a fully serverless backend on AWS that will scale with use on demand. We offload operational complexity, and ensure our backend can scale with unknown demand by choosing a serverless approach to building our platform. All of this while having a pay per use cost structure for the backend, so we’re not paying for resources that aren’t being used. The architecture enables us to rapidly deploy discrete instances of the application for development and testing purposes, and in the future it will unlock the ability to provide HeyBuddy as a white labeled dedicated platform for B2B customers.

The Serverless Architecture

Technical Highlights

Fully managed GraphQL API via AWS AppSync including real time updates

A normalised cache in a local SQLite database on the client

Integration with OpenAI and Amazon Bedrock enabling multi-model inference

DynamoDB implementing Single Table Design for performance and cost

Event Driven architecture with Amazon SNS and Amazon EventBridge

Data Lake implemented with Kinesis, AWS S3 and Amazon Athena

Defined using Cloud Development Kit (CDK), allowing consistent deployments and flexibility

The App

We designed and developed a native iOS chat app for AI companions. The robust backend implementation and intuitive frontend design enables us to offer an immersive chat experience.

The App

We designed and developed a native iOS chat app for AI companions. The robust backend implementation and intuitive frontend design enables us to offer an immersive chat experience.

An intuitive chat experience

The chat interface shares a user experience similar to other communication platforms such as Whatsapp and Telegram. All messages are updated in real time, whether the app is open or closed.

A catalogue of curated AI companions

Users can discover new companions in a swipe-style interface. Each companion has a detailed profile and is built carefully according to their unique persona.

A catalogue of curated AI companions

Users can discover new companions in a swipe-style interface. Each companion has a detailed profile and is built carefully according to their unique persona.

An AI that feels real

AI companions aren't limited to respond to messages from users. They proactively send thoughts of their own, drawing on previous dialogue. In conversation they go beyond just responding to users, towards a deeper understanding of the conversation.