10x Network Design Engineer
summary
Datacenter designs become codified rules, retrieval of hundreds of past designs, and a fleet of specialist agents that size fabrics, render configs, and generate BOMs — delivering a claimed 10x productivity speedup and turning a 6‑week sizing exercise into minutes for hundreds of analyses now deployed in production. The system uses deterministic IP addressing, capacity models, and a grounding/validation layer so agents draft real, deployable topologies but cannot yet own end‑to‑end orchestration; humans remain approvers. Expect autonomous sizing, grounded agents, and validation as the remaining hard problem.
AI-written summary · about
chapters (8)
- 0:00Introduction to AI in design
- 0:25The fragile chain problem
- 1:51Why generic LLMs fall short
- 2:53Grounding solutions
- 4:26The agent toolkit
- 5:57New workflow and production
- 7:13Practical engineering wins
- 8:37Future maturity and takeaways
related talks
30mHow We Built an Automated Data Center Network in Less Than 6 Months
16mMatryoshka: Realizing Hyperscale Data Center Network Design for the AI Era
52mThe role of the Network Engineer in the DevOps era
17mMeta’s Network Journey to Enable AI
41mHow to Build a Network Automation Stack From The Ground Up
43mArchitectural Blueprints for AI Agents
19mEngineering Facebook's Data Center Network: Challenges and Limitations
26mApplying New High-Throughput Network Solutions in IEC (Integrated Edge Cloud)