Engineer at Microsoft exploring the art and reasoning behind software systems.…
5 ratings and 7 bookings across all services
Most people approach system design by memorising architectures: “use Kafka here, add a cache there.” That breaks down the moment you face a problem you haven’t seen before, and it breaks down completely with AI agents, which are slow, expensive, non-deterministic, and call tools on their own.
I’ve spent 6 years building distributed systems, and today I work on highly scalable systems at Microsoft. I’ve recently moved into building AI agents, so I’ve seen first-hand which system design instincts still hold and which ones agents break. This session is that thinking, packaged.
What: the building blocks every system is made of, and the new ones agents add (LLM calls, tools, memory, context, orchestration)
Why: the trade-offs behind every decision: latency vs cost, consistency vs availability, and the agent-era ones like accuracy vs token cost, and autonomy vs control
How: a repeatable step-by-step framework to approach any design problem, applied live to a classic system and then to an agent-based one
Who it’s for: software engineers (1–6 years) preparing for system design interviews or starting to build with LLMs and agents
60 minutes · Live Q&A · Recording available