This repository hosts the code for a novel Goose-powered Agentic AI framework designed to revolutionize psychiatric care in India. We aim to address critical challenges in mental healthcare, including fragmented patient records, inefficient documentation, and the lack of systematic, actionable feedback for both clinicians and patients, especially within India's multilingual and multicultural context.
Our system utilizes advanced voice-based AI and an integrated digital platform to provide:
- A Comprehensive Feedback Engine for both patients and doctors.
- Automated Consultation Notes Generation for efficient documentation.
- A Full Patient Profile management system for comprehensive, longitudinal records.
- Scoring across 17 distinct communication metrics in multiple Indian languages, directly informing personalized feedback.
By leveraging an Agentic AI approach orchestrated by Goose and employing DSPy-like prompt optimization techniques, we strive to build a highly adaptive, empathetic, and culturally sensitive tool that enhances communication quality and streamlines clinical workflows in a high-stakes environment.
- Multilingual Voice Analysis: Supports 5-7 major Indian languages (Hindi, English, Bengali, Telugu, Marathi, Tamil, Gujarati) with capabilities for dialectal variation and code-switching.
- Agentic AI Architecture: Built with the Goose framework to orchestrate specialized AI agents for audio processing, linguistic understanding, notes generation, patient profile management, feedback analysis, and metric scoring.
- Prompt Optimization: Incorporates Automated prompt optimization techniques (similar to SIMBA) directly within Goose agents to ensure highly accurate, empathetic, and contextually appropriate AI outputs for notes and feedback.
- 360-Degree Feedback Engine: Provides detailed, actionable feedback to both psychiatrists and patients on communication quality, empathy, active listening, and cultural sensitivity based on 17 specific metrics.
- Automated Clinical Documentation: Generates structured, summarized consultation notes, reducing administrative burden for clinicians.
- Comprehensive Patient Profiling: Manages longitudinal patient data, addressing the critical challenge of fragmented health records and enhancing continuity of care as patients move.
- Cultural Adaptation: AI modules are designed to understand and adapt to regional dialects, cultural metaphors, family dynamics, and religious sensitivities.
- Secure Digital Platform (Mobile/Web App): A custom-built, multilingual, and secure interface for data collection, feedback delivery, and patient profile management, adhering to data protection regulations like India's DPDP Act, 2023.
- Agent Orchestration: Goose
- AI Models: India-specific foundational AI models (ASR, NLU, NLG) trained on diverse Indian linguistic datasets.
- Prompt Optimization: DSPy-like automated techniques (similar to DSPy SIMBA) for optimizing LLM interactions.
- Audio Processing: High-quality digital audio recording, noise reduction, speaker diarization.
- Speech-to-Text (ASR): Leveraging models capable of handling diverse Indian languages and accents (e.g., Bhashini, AI4Bharat's IndicWav2Vec).
- Natural Language Processing (NLP): Custom NLP models for sentiment analysis, cultural analysis, entity recognition, and summarization, fine-tuned for Indian healthcare data.
- Digital Platform: Custom mobile application (Android/iOS) and/or web interface for user interaction and data management.
- Data Storage: Secure, encrypted databases for sensitive patient and consultation data.
This project is licensed under the [Apache 2.0 License] - see the LICENSE.md file for details.
For any inquiries or collaborations, please reach out to [mohit@aximox.com] or open an issue on this GitHub repository.