Insp# KineziAI – Project Story
🔥 What Inspired Us
Both founders are physiotherapists. Every day we saw the same pattern:
physiotherapists are overloaded, patients wait months, and specialists spend 40–60% of their day on manual assessments and writing exercise programs.
Meanwhile, the market is full of fitness apps — but none that actually solve the real bottleneck:
physiotherapists don’t have time.
This inspired us to build an AI tool that automates the repetitive parts of physiotherapy while keeping specialists fully in control.
💡 What We Built
During the hackathon, we created an MVP that:
- analyzes a squat video using AI
- detects biomechanical tendencies (e.g., knee valgus, thoracic rounding)
- generates a personalized corrective exercise program
- explains why each exercise was chosen
- allows the physiotherapist to edit or customize the plan
Our MVP currently focuses on squat analysis (Phase 1).
The architecture is designed so we can later expand to posture analysis, lunge tests, mobility tests, gait, and more.
🛠 Tools & Technologies Used
- Lovable AI for rapid full-stack development
- OpenAI models for biomechanical reasoning and program generation
- Supabase for authentication and database
- Custom prompt engineering for clinical decision logic
- React/Next.js (autogenerated) for frontend
- Video input processing for movement checkpoints
Lovable helped us build a working prototype in hours, not weeks.
📚 What We Learned
- AI can automate a surprisingly large part of physiotherapy workflows — as long as humans stay in control.
- Physiotherapists value clarity and customizability more than automation alone.
- The hardest part is not generating exercises — it’s generating clinical reasoning behind them.
- A functioning MVP requires strict scope control. Focusing only on squat analysis allowed us to deliver something real.
⚠ Challenges We Faced
1. Technical: Authentication & database policies
Supabase Row-Level Security caused user profile creation issues, forcing us to debug database policies under time pressure.
2. AI Consistency
Ensuring AI-generated recommendations stayed clinically correct and not random required careful prompt design.
3. Scoped MVP Design
We had to avoid the temptation to build “everything” (posture, gait, lunges).
Choosing one assessment (squat) helped us build a stable, demo-ready MVP.
4. Time Constraints
Balancing physiotherapy expertise, AI logic, and UI clarity in less than five days required aggressive prioritization and rapid iteration.
🚀 Why We Believe This Can Become a Valuable Startup
The world doesn’t need more exercise apps — it needs scalable physiotherapy tools.
KineziAI fills the gap between:
- enterprise digital MSK solutions (Kaia, Sword), and
- individual physiotherapists who lack workflow automation tools.
The TAM is massive (“MSK” is a $60B+ market), clinicians are digitally ready, and small clinics urgently need efficiency.
Our MVP already shows that AI can meaningfully reduce workload and improve the quality of physiotherapy workflows.
With continued development, KineziAI can scale across Lithuania, the Baltics, and eventually the EU clinical market.
✅ Summary
We built KineziAI to help physiotherapists do more in less time — not to replace them, but to empower them.
The hackathon MVP proves that clinically aligned AI can automate assessments and program generation, opening the door to a scalable, impactful digital health product.iration
Log in or sign up for Devpost to join the conversation.