How to Build Software for Spies?
summary
Preligens develops on-premise AI for defense using purchased commercial satellite imagery augmented with simulated and style-transferred variants to approximate client data. They automate model discovery in an "AI Factory" made of about 120 configurable components that generate ~100,000 combinations and run ~2,000 experiments per month across 17 core architectures, iterating models every 6–8 weeks and deploying hardened offline stacks with strict licensing and predeployment testing on a bi-monthly field cadence.
AI-written summary · about
chapters (7)
- 0:00Introduction
- 6:44Complex deployment environments
- 11:39Data access and simulations
- 14:10AI model industrialization
- 20:51Logistics and deployment
- 26:53Summary and mission
- 29:01Q&A
related talks
28mSecuring AI by design
37mAGENTS OF S.E.A.L.E.D: AI Agentic Cybersecurity Framework
34mInvisible Infiltration of AI Supply Chains: Protective Measures from Adversari...
30mSecureAI: A Scalable, Secure, and Compliant AI Solution
21mEnhancing Threat Intelligence Using AI
39mSecuring AI at Scale: Practical Defenses against Prompt...
37mReinforce AI security: Protecting AI applications, models, and data
14mAI Sprawl: The Hidden Risk Threatening Your Enterprise Now