CBT Advice
⬇ 4 ♥ 0
t5-cbt-synth
Built with synthetic data using local models to iterate and control output. Designed to generate Cognitive Behavioral Therapy (CBT) advice.
Output: Therapeutic guidance
Portfolio
AI Models
These generative AI models were created, trained, and fine-tuned by me to detect emotion and respond effectively to human statements. Hosted on Hugging Face.
Key Skills Demonstrated:
Generative AI Fine-Tuning Synthetic Data Generation Vector Embeddings & Semantic Search Emotion Detection Data Curation & Engineering Local Model Iteration
Emotion Detection
⬇ 0 ♥ 0
empathetic-deberta-v3
Emotion detection model that identifies the emotional state from text. Returns a single emotion label (e.g., "anxious", "happy", "frustrated"). Trained on curated therapy notes using embeddings and Euclidean distance.
Output: Single word (emotion label)
Therapy Response
⬇ 0 ♥ 0
empathetic-t5
Therapy response model trained to generate empathetic, supportive replies. Provides thoughtful responses similar to a therapist. Built using curated therapy notes and semantic similarity filtering.
Output: Empathetic reply