Many crew-related errors in aviation are caused by hazardous cognitive states of the crew including overstress, disengagement, high fatigue and ineffective crew coordination. Aviation safety can be improved by monitoring and predicting these cognitive states in a non-intrusive manner and designing mitigation strategies. In Next Generation Air Transportation System (NextGen) flight deck, there will be a transition from ground based navigation infrastructure to satellite based navigation and some control of separation of traffic will be delegated to the cockpit from Air Traffic Control (ATC). This will increase responsibilities of pilots, making hazardous cognitive state assessment more critical. To address this, IAI and its collaborators, the University of Iowa and Old Dominion University have been awarded a contract entitled, “Crew Systems Technologies for Improved Aviation Safety.” A real-time hazardous pilot Cognitive State Assessment system, called CSA-Deep, will be built for Integrated Crew-System Interaction (ICSI) in all phases of flight. The key innovation here is the modeling and adaptive updating of hazardous cognitive states using a large amount of unlabeled and limited labeled data through semi-supervised deep learning. A unified multiple cognitive states assessment framework will be built based on multiple non-intrusive sensing modalities including eye tracking information, voice, heart and respiration rates, body temperature and posture. Hazardous labels for each cognitive state will be output by an Integrated Alerting and Notification (IAN) system. This research will leverage ongoing flight deck studies being conducted by this team, including operator functional state assessment under NASA SBIR Phase II, OSD/Army voice-based stress analysis Phase II, and NextGen work under the NRA.