AI mental health platform
TheraAI
A two-sided mental health platform. Patients check in, journal and talk to an AI companion between sessions, and their psychiatrist walks into every appointment already briefed.
- Industry
- Healthcare
- Our role
- Design and full-stack development
- Timeline
- 3–6 months
- Platform
- Web
The goal
Help people notice how they're doing between appointments, and give their psychiatrist that picture before every session.
The challenge
Two very different audiences on one backend, some of the most sensitive data there is, and an AI that must never be the only one handling a crisis.
How we solved it
A calm patient app and a focused psychiatrist app on one FastAPI backend, with PyTorch emotion models running on every journal entry, records visible only to the psychiatrist a patient booked, and crisis language routed straight to a clinician.
What we built
Check-ins in one tap
Eight moods, one tap. Each check-in becomes a cell on a 30-day band, so patterns show up on their own.
Live emotion analysis
PyTorch models tuned to run inside FastAPI, so every journal entry is analysed as it saves.
A companion that remembers
Keeps context between conversations, takes voice input, and writes a weekly read on how things are going.
Validated screenings
PHQ-9, GAD-7 and PSS-10, scored instantly and kept so patients and clinicians can compare over time.
A workspace for psychiatrists
Caseload dashboard, pre-session briefings, patient records, treatment plans and Google Calendar sync.
Safety by design
Crisis language pauses the companion, shows local helplines and alerts the psychiatrist. Scores start a conversation, never a diagnosis.
One tap is enough.
Logging a mood takes one tap. Each one becomes a cell on a 30-day band, so patterns show up without any effort.
It listens, and it remembers.
A chat companion that keeps context between conversations, plus reflections on every journal entry and a weekly read on how things are going.
From self-check to a real psychiatrist.
Validated screenings scored instantly and kept for comparison, then video sessions booked inside the app.
Walk in already knowing how the fortnight went.
A caseload dashboard with each patient's latest mood, the next sessions, and a briefing before every appointment.
One record for every patient.
Sessions, notes, prescriptions, exercises and the treatment plan sit together, and the calendar syncs with Google Calendar.
It steps aside when it should.
If a chat or journal entry shows crisis language, the companion stops, shows crisis contacts, and the psychiatrist gets an alert on their dashboard.
A quieter place to start.
The landing page shows the product instead of describing it: a sample journal entry, the feelings picked up in it, and eight weeks of mood on one line.
What shipped
- Patient app with mood check-ins, a journal and history
- Psychiatrist app with caseload, briefings and records
- AI companion with memory and voice input
- Real-time emotion analysis on every journal entry
- PHQ-9, GAD-7 and PSS-10 screenings
- Video sessions booked inside the app
- Treatment plans, prescriptions and exercises
- Google Calendar sync
- Crisis alerts routed to the psychiatrist
- Landing page, plus light and dark themes
Services
Stack
- React
- FastAPI
- PyTorch
- MongoDB
- Google Calendar API
Building something like TheraAI?
Tell us what you need and we'll tell you honestly how we'd build it.
Book a free callNext project
NexDash
Private seller dashboard