Inspiration
1 in 4 people with bipolar disorder go undiagnosed for over a decade. The current tools can be expensive, and too slow to catch the rapid shifts that define the condition. We wanted to build something that could aid people facing Bipolar feelings that is easily accesible.
What it does
Users complete a 60-second voice check-in answering four questions. Anchor analyzes their vocal prosody, speech rate, and word content using AI to generate bipolar state indices. The home screen surfaces their energy, mood, and arousal trends in plain language. A medications tracker keeps doses on schedule, and a onboarding captures each user's personal episode symptoms.
How we built it
React Native + Expo for cross-platform mobile Hume AI for real-time vocal emotion analysis, energy, valence, arousal, and top detected emotions Anthropic Claude to interpret transcripts and generate bipolar indices, speech organization scores, and content theme flags Firebase Firestore + Realtime Database to store every check-in, medication log, and conversation
Challenges we ran into
Getting meaningful signal from short voice responses was the hardest part. 60 seconds of audio is noisy. We had to carefully engineer the Claude prompt to extract clinically relevant data. Syncing Hume's async audio analysis with Claude's transcript interpretation required a precise pipeline. We also had to design the UX so it felt supportive, not clinical or alarming.
Accomplishments that we're proud of
Generating a real bipolar state index, manic, depressive, mixed, from a single voice check-in. The fingerprint onboarding that personalizes detection to the individual's own symptoms is something no consumer app does today. We're also proud of how the app feels.
What we learned
Emotion AI alone isn't enough. Valence and energy scores mean very little without the linguistic layer on top. What someone says matters as much as how they say it. We also learned that designing for mental health requires extreme care in how results are framed.
What's next for Anchor
Our immediate next step is shipping the MVP, validating the core voice check-in feature with real users, and building a social media presence to grow awareness. From there, we're pursuing partnerships with DBSA and NAMI to lend clinical credibility, launching in-app monetization that keeps the app free for users, and attending local mental health conferences like NAMI Eastside YMCH to connect with the communities we're building for.
Built With
- anthropic
- antigravity
- claude
- eleven-labs
- firebase
- hume
- react-native
Log in or sign up for Devpost to join the conversation.