Barry’s web interface and application logic are built on Next.js, delivering a responsive, high-contrast black, white, and orange interface. We integrated Supabase for user authentication and relational/vector storage, creating a secure, persistent memory vault that locks in a user's medical history, accommodations, and communication preferences after a single briefing. To power live appointment mode, we integrated Amazon Transcribe to handle high-accuracy, low-latency clinical speech-to-text. This allows Barry to listen to rapid doctor-patient conversations in real time, bridging the communication speed gap for non-verbal users and those with speech impairments. The core advocacy engine relies on Claude hosted via Amazon Bedrock, orchestrating complex contextual reasoning, real-time response generation, and post-visit insurance appeal automation. Development was accelerated using Codex and OpenCode for agent workflow automation, code generation, and rapid iteration during the hackathon.

Balancing low-latency streaming between Next.js, Amazon Transcribe, and Amazon Bedrock was a major hurdle. In live clinical settings, non-verbal patients need near-instantaneous advocacy generation without sacrificing Claude’s medical reasoning accuracy. Structuring Supabase memory persistence to carry over critical accommodation needs without cluttering the context window with irrelevancies from past appointments required careful prompt engineering and state management in Next.js. Designing a modern interface with a black, white, and orange palette while adhering strictly to high-contrast WCAG AAA accessibility standards meant carefully tuning accent colors and typography so screen readers and visually impaired users experience zero visual fatigue.

Designing for non-verbal and chronically ill users forced us to build cleaner Next.js component architectures, tighter API payloads, and far more intuitive user flows than a standard healthcare app would ever require. Moving away from single-session chat models to persistent agentic companions completely eliminates repetition fatigue. One briefing changes how patients experience healthcare forever. AI in healthcare should never act as a black-box replacement for doctors or patients; it must serve as an empowered, user-controlled advocate that amplifies the patient's own voice.