SomaAI - Pain Awareness & Relief Aid

Chronic pain affects 1 in 5 people worldwide - over 50 million in the U.S. alone. SomaAI provides accessible AI-guided support, inspired by Pain Reprocessing Therapy, to ease suffering.

Comment

Inspiration

This project started from a deeply personal place. My mum in Nigeria has lived with chronic pain for years. Almost every older person I know back home talks about pain as though it is inevitable. Seeing my mum struggle to get through daily life and hearing the same conversations among relatives made me determined to find a way to help.

When I was accepted into my university’s (CSUMB) UROC research program, I was paired with Professor Michael Guerrero, who has also lived with chronic pain. His guidance and encouragement, along with my family’s experience, inspired me to merge my interest in AI and interactive 3D technology with a project that could support real people.

That search led me to Pain Reprocessing Therapy (PRT). What stood out was the scientific evidence. A 2021 randomized controlled trial from the University of Colorado showed that after four weeks of PRT, two-thirds of participants were nearly or completely pain free. Brain imaging confirmed reduced activity in pain processing regions, showing that the therapy is not just psychological but also neurological. These proven results gave me confidence that building an AI tool around PRT could have meaningful impact.

What it does

SomaAI is an AI-powered pain awareness and relief aid inspired by PRT.

Users log their pain by selecting where it hurts, rating severity, describing the quality, and noting how long it has persisted.

The system sends this information to a fine-tuned GPT-OSS model trained on curated PRT examples.

The AI responds with short, supportive coaching messages and practical exercises designed to help users practice somatic tracking.

Every response ends with a clear disclaimer that it is not medical advice. Escalation rules trigger caution messages if red flags are detected.

How I built it

Frontend: Unity, allowing rapid prototyping of interactive 3D elements like the anatomy figure.

Backend: Python with FastAPI and Uvicorn, connected to Hugging Face’s GPT-OSS inference API.

Model: GPT-OSS, fine-tuned with curated data points from PRT research and transcripts to improve tone, structure, and relevance.

Integration: Unity communicates with the Python backend via HTTP/JSON.

Safety: Built-in disclaimers and escalation triggers for severe or risky cases.

Challenges I ran into

  • The main challenge I had was being able to consistently get output from the model. Sometimes the model gave me empty strings.

  • Parsing outputs: Hugging Face’s variable JSON formats caused truncated replies. Fixed with stop tokens and robust parsing.

  • Training data: Needed careful curation to keep examples safe and evidence-inspired.

  • Tone balance: Designing responses that feel supportive without sounding like clinical advice.

  • Deployment: Making the backend easy for non-technical testers — solved with one-click .bat and .sh scripts.

Accomplishments that I am proud of

  • Built a full end-to-end system: Unity app, backend, and GPT-OSS integration.

  • Developed a safe prompting framework with disclaimers and escalation rules.

  • Created an interactive body map that makes pain logging intuitive.

  • Began fine-tuning GPT-OSS with PRT data for smarter, evidence-aligned responses.

  • Turned personal inspiration into a working demo that could genuinely help people like my mum and my mentor.

What I learned

  • How to connect Unity with a Python backend and AI models.

  • The importance of prompt engineering and stop conditions to keep responses safe and complete.

  • That human-centered design is essential for health-related tools.

  • How evidence-based research like PRT can be translated into accessible AI-powered coaching.

What’s next for SomaAI — Pain Awareness & Relief Aid

  • Expand fine-tuning to 10,000+ PRT data points for smarter coaching.

  • Add retrieval-augmented generation (RAG) so the model cites evidence-based sources.

  • Extend the Unity app with richer 3D experiences and potential VR scanning support.

  • Partner with research labs or clinicians to validate SomaAI in controlled studies.

  • Release mobile versions to reach communities like the one my mum is part of in Nigeria, where access to pain specialists is limited.

✨ SomaAI combines science and lived experience. It is my way of using technology, research, and compassion to reduce suffering and bring relief to millions living with chronic pain.

Built With

Share this project:

Updates