Metis pioneers AI and ML to transform aerospace operations, leveraging expertise from NASA’s ARTS Program to deliver cutting-edge solutions in air traffic management, urban air mobility, environmental forecasting, and supply chain resilience.
- Natural Language Processing (NLP): Automating aviation communications with NLP models for transcription, intent recognition, and slot filling, enhancing air traffic control efficiency.
- Reinforcement Learning: Optimizing airspace deconfliction for autonomous UAM operations, enhancing safety in dynamic environments.
- Large Language Models (LLMs): Powering intelligent decision-making in ATC and automation with fine-tuned LLMs, ensuring transparency and safety.
- Predictive Modeling: Forecasting airport operations with ML models for runway assignments and resource allocation, validated for accuracy.
Key AI applications include:
- ALFRD - Aviation LLM: Metis developed a large language model for air traffic control, processing over 2,000 Digital Taxi instructions with explainable AI methods (Layer-wise Relevance Propagation and Kernel Shap).
- UADT – UAM Digital Twin: Applies reinforcement learning and LLMs to optimize deconfliction and conformance monitoring for small Unmanned Aircraft Systems (UAS) beyond visual operations, simulating 25,000+ scenarios.
- Sherlock – Data Warehouse Platform: Developed an enterprise platform for NAS-wide ATM data collection, processing, and delivery, leveraging machine learning to analyze flight, weather, traffic management, and facility data from FAA System Wide Information Management (SWIM) and other sources for predictive and operational insights.
- PrimeGov – Supply Chain: Integrated a large language model for enhanced supply chain data search and analysis, supporting NASA and Defense Contract Management Agency (DCMA) operations.