AI Mental Health Chatbot
Developed a fine-tuned conversational AI using BlenderBot for empathetic and safe mental health dialogue. Built on the MentalChat16K dataset with custom preprocessing, evaluation, and deployment workflow
MindEase: AI Mental Health Chatbot
MindEase is an AI-powered conversational chatbot designed to provide compassionate dialogue and emotional support through fine-tuned generative AI on real-world mental health data.
Fine-tuned Meta’s BlenderBot-400M-distill model on the MentalChat16K dataset, integrating text preprocessing, tokenization, and conversational context optimization. Created train/validation/test splits and automated preprocessing pipelines to improve data quality.
The model needed to maintain empathy and coherence across sensitive topics while operating efficiently on CPU. Balancing response diversity with safety and accuracy required careful fine-tuning and evaluation strategy.
Deployed a CPU-optimized BlenderBot fine-tuning process using Hugging Face Transformers, integrated evaluation (ROUGE-L, F1), and wrapped the model with a FastAPI backend. Built a frontend chat interface with HTML/CSS/JS and served via API.
Key Features
- Fine-tuned BlenderBot on MentalChat16K
- Emotionally adaptive responses
- FastAPI backend integration
- Hugging Face model deployment
- Lightweight CPU inference
- Web-based chat frontend