Project Name:
Accelerating Intervention of Military Self-Harm Using AI Data-Driven Technology
Organization Name and Division:
Air Force, Department of Defense, in collaboration with Department of Veterans Affairs, Amazon Web Services (AWS), Booz Allen, RallyPoint, and Harvard University
Year and Quarter: 2021- 2022
DoD SBIR x19.2 AF193-DCS01 Contract:
FA8649-20-9-9128
NAICS Codes:
541715 – Research and Development in the Physical, Engineering, and Life Sciences
Overview
Launch Theory, working as a subcontractor under Booz Allen Hamilton, played a vital role in a Department of Defense SBIR project aimed at preventing military suicide through early intervention. The project brought together an elite team including the Department of Veterans Affairs, Harvard University’s Nock Lab, and Amazon Web Services (AWS) to develop an AI-powered intervention system for detecting early warning signs of self-harm in social media communications.
Our Challenge
Upon initial assessment, Launch Theory identified ten years of technical debt that would have hindered any effective AI integration. This discovery led to a complete architectural overhaul of the planned system.
Project coordination presented unique challenges, requiring Launch Theory to manage over 60 individuals across multiple organizations, including AWS AI modeling scientists, Harvard researchers, VA representatives, and various technical teams. Launch Theory established clear communication channels, managed regular cross-team meetings, and ensured all stakeholders remained aligned with project objectives while meeting aggressive timelines.
What We Accomplished
Cloud Architecture Development: Launch Theory implemented a sophisticated AWS-based message processing system architecture that allowed the aging social network to utilize AI processing despite the existing technical debt. The team developed a separate cloud-based architecture that connected to the aging system through custom, secure, encrypted API endpoints in both systems. This innovative approach enabled high-volume batch processing of messages without security risks while also meeting strict SOC 2 compliance requirements. The system’s robust design ensured seamless integration with existing operations while providing a foundation for future scalability.
Privacy and Security Implementation: Recognizing the critical nature of data protection for service members, Launch Theory developed a sophisticated double-blind GUID system that maintained user anonymity while enabling effective intervention. The team created a custom data anonymization engine that automatically stripped all personally identifiable information (PII) and location data from messages, ensuring that even in the event of a breach, sensitive data would remain protected. This comprehensive security architecture established new standards for protecting service member privacy while maintaining system functionality.
AI/ML Integration: The team implemented a state-of-the-art machine learning infrastructure integrated with Amazon SageMaker and Amazon Comprehend. This system incorporated specialized pattern recognition for military-specific language and included a human-in-the-loop validation system for continuous model refinement. The real-time intervention delivery system enabled immediate response to identified risks while providing an interactive system that gradually increased scoring accuracy standards. At the time, this technical achievement represented a significant advancement in applying AI technology to mental health support.
“All of us at Launch Theory were incredibly proud to be a part of this vital initiative,” notes Greg Stout, Launch Theory’s founder. “Our company excels at bridging the gap between innovative technology and mission-critical needs. Working alongside the Department of Veterans Affairs, Harvard University, and AWS, we successfully implemented a groundbreaking secure AI system focused on supporting the mental health of the men and women who have served this nation. Collaborating with visionaries like Dr. Nock and seeing how our technology skill could directly impact service members’ lives has deeply moved our entire team. This experience has strengthened our commitment to developing solutions that make a real difference in the lives of those who serve our country.”
Greg Stout, Founder and CEO Launch Theory
Project Timeline: 15 Months
Phase 1
Infrastructure Design and Security Setup
Phase 2
AI Model Integration and Testing
Phase 3
Deployment and Ongoing Monitoring

Services Used
- Technical Leadership and Strategic Consulting
- AI and Machine Learning Integration
- Technical Debt Assessment
- Cloud System Architecture Development (AWS)
- AWS Infrastructure Development
- Custom Data Anonymization & Encryption Tools
- Secure API Gateway Implementation
- Data Security & Encryption Management
- Compliance with DoD SOC 2 Standards
- Quality Assurance & Regression Testing
- Performance Optimization
- Project Management & System Engineering
- Accessibility Compliance (508, WCAG)

Key Technologies
- Amazon Web Services (AWS) Cloud Infrastructure, S3, EC2, Lambda,
- Machine Learning Models (AI/ML), SageMaker, Comprehend
- Custom Geo-Targeted Intervention Messaging Systems
- Custom Data Encryption and Anonymization Tools
- Double-Blind GUID System for Data Privacy
- Ruby on Rails

Deliverables
- Scalable infrastructure for real-time data processing
- AI models of Harvard’s suicide risk prediction algorithms
- Complex System Documentation for Security Audits
- Continuous monitoring and compliance reporting tools
Results
The system’s performance demonstrated remarkable improvements throughout the pilot:
- Achieved 8x increase in processing speed from the original prototype
- Facilitated approximately 6,000 interventions
- Improved pattern-matching accuracy by 500%
The year-long pilot demonstrated clear success in its scientific objectives, supporting Dr. Nock’s hypothesis that technology could help reach service members who “don’t seek out traditional mental health care.” (Federal News Network, 2023) The project established several key operational achievements:
- Successfully processing interactions from over 2 million users in near real-time
- Maintained 100% system uptime, even during pilot phase
- Completed all DoD SOC2 security audits with zero critical findings
- Met all DoD compliance requirements
This project exemplifies Launch Theory’s ability to serve as both technical advisor, technical implementer and project coordinator for complex, multi-stakeholder initiatives while delivering sophisticated technical solutions that address critical social challenges. By successfully proving the viability of AI-driven early intervention, the project laid crucial groundwork for future innovations in suicide prevention technology while maintaining the highest standards of security and privacy protection.