Sakhi: An AI-Powered Mental Health Companion

Inspiration

Mental health support is often inaccessible due to stigma, cost, or lack of available resources. We wanted to create a solution that provides personalized, empathetic, and always-available assistance. Inspired by the need for a safe, judgment-free space, Sakhi was born—a compassionate AI-powered companion designed to support users through their mental health journey.

What it does

Sakhi is an AI-driven mental health companion that:

  • Engages users in meaningful conversations to provide emotional support.
  • Helps users track their mood and well-being over time.
  • Uses LLM-based responses fine-tuned for empathy and context.
  • Supports voice-based interaction using text-to-speech and speech-to-text.
  • Maintains user privacy with secure data handling.
  • Offers therapeutic techniques like guided journaling and breathing exercises.

How we built it

We developed Sakhi using a combination of:

  • Gemini (RAG) for generating contextual and empathetic responses
  • Blender for embedding image-based interaction
  • ReactJS for a seamless and interactive user experience
  • Text-to-Speech & Speech-to-Text APIs to enable voice-based conversations
  • Secure Backend & Data Handling to ensure privacy and confidentiality

The system was designed to provide real-time guidance, mood tracking, and conversational therapy, making mental health support more approachable and engaging.

Challenges we ran into

  • Balancing empathy & accuracy: Ensuring the AI provided supportive yet responsible advice without overstepping its boundaries.
  • Real-time response optimization: Fine-tuning the LLM while maintaining efficiency was a challenge.
  • Ethical considerations & data privacy: Implementing strong security measures to protect user data and ensure compliance with mental health regulations.

Accomplishments that we're proud of

What we learned

  • Fine-tuning LLMs for empathetic and personalized responses is crucial for mental health applications.
  • User experience plays a significant role in engagement—creating an intuitive and friendly interface was key.
  • Real-world testing is essential; feedback from mental health professionals and users helped refine the interactions.

What's next for Sakhi

We aim to integrate multi-lingual support, expand personalized mental health plans, and explore partnerships with professionals to make Sakhi a more comprehensive and effective mental health companion.

🌍 Beyond Mental Health: AI Agents for Multiple Use Cases

Our work on Sakhi opens up opportunities to develop AI-driven agents for various real-world applications. In the future, we can create AI agents for:

  • Conducting interviews and evaluating candidates.
  • Personalized learning assistants for education.
  • Customer support bots with advanced conversational abilities.
  • Healthcare chatbots for patient guidance.

Built With

Share this project:

Updates