How Revolut Cut AI Coding Costs 90% With Self-Hosted Open Models 31 October 2026·5 mins AI Engineering AI Infrastructure Talk Notes Ai-Engineering Llm-Inference Open-Models Self-Hosting Coding-Agents Glm Developer-Productivity Fintech
Mixture-of-Kittens: MoE Megakernel for NVL72s 3 October 2026·11 mins AI Infrastructure Distributed Systems Paper Notes Distributed-Systems Arxiv Mixture-of-Experts Gpu-Kernels Llm-Training Nvlink
From LLM Inference to Agentic Workloads: Characterization and Implications for Serving Systems 22 August 2026·12 mins AI Infrastructure Operating Systems Paper Notes Operating-Systems Arxiv Llm-Serving Agents Systems
Who Should Own the Expert Cache? Kernel-Managed Tiering for Trillion-Parameter MoE Inference 15 August 2026·9 mins AI Infrastructure Operating Systems Paper Notes Operating-Systems Arxiv Mixture-of-Experts Llm-Inference Page-Cache Systems
On Topology's Role in ML Training Performance 8 August 2026·9 mins AI Infrastructure Networking Paper Notes Networking Arxiv Distributed-Training Interconnect Collectives
Architectural Implications of Agentic AI Workflows 8 August 2026·8 mins AI Infrastructure AI Engineering Paper Notes Distributed-Systems Arxiv Agents Datacenter Llm-Serving
Incast-Free MoE Rate-Based Scheduling 1 August 2026·10 mins AI Infrastructure Networking Paper Notes Networking Distributed-Systems Arxiv Mixture-of-Experts Datacenter Congestion-Control
Windowed-MTP: Removing the Full-Context Draft-KV Tax at Million-Token Context 25 July 2026·3 mins Machine Learning AI Infrastructure Paper Notes Machine-Learning Arxiv Inference Speculative-Decoding Long-Context Llm-Systems
Who Needs DRAM? We Have Fiber 11 July 2026·8 mins AI Infrastructure Operating Systems Paper Notes Distributed-Systems Arxiv Systems Optics Llm-Infra Memory
ARGUS: Production-Scale Tracing and Performance Diagnosis for 10,000+ GPU Clusters 20 June 2026·9 mins AI Infrastructure Distributed Systems Paper Notes Distributed-Systems Arxiv Gpu Observability Llm-Training Fail-Slow
AgileOS: A GPU Operating System Layer for Protected CUDA Services 13 June 2026·4 mins Operating Systems AI Infrastructure Paper Notes Systems Gpu Operating-Systems Cuda Arxiv
Inside Jane Street's GP300 Training Data Center 23 May 2026·7 mins AI Infrastructure Networking Talk Notes Systems Infrastructure Gpu Datacenter Ai-Infrastructure Liquid-Cooling Networking