Detect Malicious Attacks at 84M RPS with ML Under 500µs

Learn how to architect and optimize distributed systems for detecting malicious attacks at massive scale through this conference talk from LeadDev Berlin 2025. Explore Cloudflare's high-performance architecture that processes over 84 million requests per second while maintaining sub-millisecond latencies for threat detection. Discover the distributed system design spanning global data centers and understand how Rust programming language enables extreme efficiency and resource optimization where every CPU cycle and memory byte matters. Examine critical smart processing techniques including advanced caching strategies and hardware tuning that enable success at this scale. Dive deep into the machine learning models powering Cloudflare's Web Application Firewall (WAF) and learn about extensive optimizations including SIMD operations, TensorFlow Lite upgrades, and LRU caching implementations that achieve incredible speed. Understand innovative data generation and sampling strategies essential for training accurate and resilient AI models in production environments. Gain practical insights into performance optimization techniques, challenges of deploying machine learning models in high-stakes real-time environments, and actionable strategies for leveraging serverless and distributed architectures to scale your own applications and ML algorithms.