Coursera Flash Sale
40% Off Coursera Plus for 3 Months!
Grab it
Explore how large language models have fundamentally disrupted traditional application infrastructure assumptions in this 14-minute conference talk. Learn why LLMs' inherent unpredictability and latency have rendered conventional web infrastructure patterns obsolete, as p99 latencies that once triggered alerts have become standard p1 performance metrics. Discover the infrastructure challenges facing developers building LLM-powered applications, including the limitations of current serverless providers that can't accommodate the extended execution times required by AI agents. Examine critical infrastructure patterns that have evolved from niche solutions to essential requirements for LLM applications, including durable execution frameworks, long-running workflow management, persistent API architectures, and agent-scoped storage solutions.
