Modern aviation has integrated AI into various systems, yet there remains a crucial gap where AI-driven assistance can enhance pilot decision-making. Leveraging Generative AI and Large Language Models (LLMs), we developed MAYDAY, an AI-powered tool designed to assist pilots in real-time, reducing stress and improving situational awareness.
- Uses a Retrieval-Augmented Generation (RAG) approach.
- Retrieves data from aircraft manuals, airline SOPs, and operational documents.
- Generates context-aware checklists for both standard and emergency situations.
- Supports manual text input and speech-to-text functionality for ease of use in real-world cockpit environments.
- Pilots can describe minor system failures via speech input or manual text.
- MAYDAY predicts potential cascading failures and provides early warnings to help pilots take proactive measures.
- Offers a real-time 3D visualization of the aircraft’s approach glide slope.
- Detects deviations and triggers go-around alerts if the aircraft overshoots the ideal descent path.
- Integrates real-time weather forecasts into flight planning.
- Assists in selecting alternate airports in case of adverse weather conditions.
- LLM Generation: Open-source Hugging Face SLMs with remote inference via Mistral-7B-Instruct-v0.3.
- Backend: Python (FastAPI), FAISS for retrieval.
- Frontend: Streamlit-based UI for pilot interaction.
- Speech-to-Text: Integrated voice input for hands-free interaction.
- Deployment: AWS EC2 instance for hosting.
- Download & Unzip the repository.
- Edit .env file to icnlude your huggingface API key
- Install Dependencies:
pip install -r requirements.txt python login.py