Inspiration
We were struck by a single, painful realization blood donation systems are designed for efficiency, not empathy. For Thalassemia patients, this means blood may arrive, but compassion often doesn't.
What it does
SoulBlood is an AI-powered platform that connects blood donors to Thalassemia patients not just by blood type — but by empathy. Core Features:
Emotion-AI Donor Matching: Donors upload a short voice note or answer a few emotional prompts. A Siamese BERT model matches their emotional profile with anonymized patient stories that resonate on a human level.
Heartprint Generator: After donating, donors receive an AI-generated emotional visualization — a symbolic dreamscape made using GANs — showing the “life they helped power.”
Patient Sentiment Tracker: A DL-powered model processes patient journals or voice logs to alert caregivers of emotional distress, ensuring mental health care doesn’t get sidelined.
How we built it
We used a combination of multimodal deep learning, creative storytelling, and ethical design principles. Siamese BERT for emotion-based donor-patient matching. wav2vec 2.0 for voice emotion analysis. StyleGAN2 to generate personalized "Heartprints". Transformer-based sentiment classifier for monitoring patient well-being. Built using Python, PyTorch, HuggingFace Transformers, React, Firebase, and FastAPI.
Challenges we ran into
Multimodal fusion: Combining emotion signals from voice, text, and optional face data was complex and required careful preprocessing.
Ethical storytelling: Creating an emotionally powerful system without exploiting patient vulnerability demanded strict design guidelines (anonymization, consent layers).
Synthetic dataset creation: With limited access to real patient emotional data, we had to simulate diverse and realistic emotional profiles.
Maintaining realism in GAN outputs: Ensuring AI-generated art felt human without overstepping into false representation was a nuanced challenge.
Accomplishments that we're proud of
Created an emotionally intelligent AI system that goes far beyond standard ML applications.
Successfully integrated Siamese NLP, GANs, and sentiment models into a single pipeline.
Developed a proof-of-concept that turns compassion into code, and back into human action.
What we learned
AI isn’t just for automating it can also be for amplifying humanity. Models like BERT and wav2vec can be adapted for empathetic computing, not just classification tasks. Designing for emotion is harder and more important than designing for engagement. With the right framing, deep learning can spark deeper caring
What's next for SoulBlood
Partner with Blood Warriors and NGOs to collect real anonymized patient stories, with proper consent protocols. Improve the Heartprint engine with more personalized animation and emotion-driven prompts. Launch a WhatsApp-based version of SoulBlood for rural outreach. Train the system for multilingual emotion matching in Hindi, Tamil, Bengali, etc. Pilot test in Thalassemia centers across 2 Indian cities by the end of the year.
Built With
- api
- cnns
- hugging-face-transformers-(bert
- natural-language-processing
- python
- react.js
- transformer-based
- wav2vec-2.0)
