<?xml version="1.0" encoding="UTF-8"?>
<urlset xmlns="http://www.sitemaps.org/schemas/sitemap/0.9" xmlns:image="http://www.google.com/schemas/sitemap-image/1.1">
<url>
<loc>https://inferenceengineering.co</loc>
<lastmod>2026-11-01T14:21:15.925Z</lastmod>
<changefreq>daily</changefreq>
<priority>1</priority>
</url>
<url>
<loc>https://inferenceengineering.co/sections/kv-cache-systems</loc>
<lastmod>2026-11-01T14:21:15.925Z</lastmod>
<changefreq>weekly</changefreq>
<priority>0.5</priority>
</url>
<url>
<loc>https://inferenceengineering.co/sections/ai-storage</loc>
<lastmod>2026-10-26T12:06:36.955Z</lastmod>
<changefreq>weekly</changefreq>
<priority>0.5</priority>
</url>
<url>
<loc>https://inferenceengineering.co/sections/inference-serving</loc>
<lastmod>2026-10-25T13:10:47.882Z</lastmod>
<changefreq>weekly</changefreq>
<priority>0.5</priority>
</url>
<url>
<loc>https://inferenceengineering.co/sections/gpu-memory</loc>
<lastmod>2026-10-14T12:46:16.624Z</lastmod>
<changefreq>weekly</changefreq>
<priority>0.5</priority>
</url>
<url>
<loc>https://inferenceengineering.co/sections/high-speed-interconnects</loc>
<lastmod>2026-10-30T12:31:23.337Z</lastmod>
<changefreq>weekly</changefreq>
<priority>0.5</priority>
</url>
<url>
<loc>https://inferenceengineering.co/sections/features</loc>
<lastmod>2026-09-18T06:51:30.742Z</lastmod>
<changefreq>weekly</changefreq>
<priority>0.5</priority>
</url>
<url>
<loc>https://inferenceengineering.co/authors/priya-subramaniam</loc>
<lastmod>2026-11-01T14:21:15.925Z</lastmod>
<changefreq>weekly</changefreq>
<priority>0.4</priority>
</url>
<url>
<loc>https://inferenceengineering.co/authors/amara-osei-bonsu</loc>
<lastmod>2026-11-01T14:21:15.925Z</lastmod>
<changefreq>weekly</changefreq>
<priority>0.4</priority>
</url>
<url>
<loc>https://inferenceengineering.co/authors/tobias-wrenfeld</loc>
<lastmod>2026-11-01T14:21:15.925Z</lastmod>
<changefreq>weekly</changefreq>
<priority>0.4</priority>
</url>
<url>
<loc>https://inferenceengineering.co/authors/nikolai-vereshchagin</loc>
<lastmod>2026-11-01T14:21:15.925Z</lastmod>
<changefreq>weekly</changefreq>
<priority>0.4</priority>
</url>
<url>
<loc>https://inferenceengineering.co/posts/kv-cache-offloading-stack-decisions-from-gpu-dram-to-object-storage</loc>
<image:image>
<image:loc>https://mcxtywzsmeezbwapdahq.supabase.co/storage/v1/object/public/canvas-images/1ea5b494-0bbf-4b2b-b155-3eecc768ddf1/canvas/f21f0e25-0dd0-49e2-b6aa-ab4ef4acca9d/generated/941ecff2-424a-45ca-ab2d-b2119a28bd0d.jpeg</image:loc>
</image:image>
<lastmod>2026-11-01T14:21:15.925Z</lastmod>
<changefreq>monthly</changefreq>
<priority>0.7</priority>
</url>
<url>
<loc>https://inferenceengineering.co/posts/controlling-kv-cache-memory-growth-without-sacrificing-throughput</loc>
<image:image>
<image:loc>https://mcxtywzsmeezbwapdahq.supabase.co/storage/v1/object/public/canvas-images/1ea5b494-0bbf-4b2b-b155-3eecc768ddf1/canvas/f21f0e25-0dd0-49e2-b6aa-ab4ef4acca9d/generated/b85575ce-b6f8-4d49-bf25-e062c815e55b.jpeg</image:loc>
</image:image>
<lastmod>2026-10-31T12:51:21.480Z</lastmod>
<changefreq>monthly</changefreq>
<priority>0.7</priority>
</url>
<url>
<loc>https://inferenceengineering.co/posts/kv-cache-transfer-over-rdma-between-prefill-and-decode-nodes</loc>
<image:image>
<image:loc>https://mcxtywzsmeezbwapdahq.supabase.co/storage/v1/object/public/canvas-images/1ea5b494-0bbf-4b2b-b155-3eecc768ddf1/canvas/f21f0e25-0dd0-49e2-b6aa-ab4ef4acca9d/generated/103ff2cd-b02f-4e9d-ba66-3e01cf06e43c.jpeg</image:loc>
</image:image>
<lastmod>2026-10-30T12:31:23.337Z</lastmod>
<changefreq>monthly</changefreq>
<priority>0.7</priority>
</url>
<url>
<loc>https://inferenceengineering.co/posts/rocev2-congestion-control-for-ai-storage-traffic</loc>
<image:image>
<image:loc>https://mcxtywzsmeezbwapdahq.supabase.co/storage/v1/object/public/canvas-images/1ea5b494-0bbf-4b2b-b155-3eecc768ddf1/canvas/f21f0e25-0dd0-49e2-b6aa-ab4ef4acca9d/generated/f5c414a2-e73a-402b-b7f3-cd7d4d75ff11.jpeg</image:loc>
</image:image>
<lastmod>2026-10-29T12:56:20.251Z</lastmod>
<changefreq>monthly</changefreq>
<priority>0.7</priority>
</url>
<url>
<loc>https://inferenceengineering.co/posts/how-rdma-bypasses-the-cpu-to-sustain-all-reduce-throughput-in-gpu-cluster-traini</loc>
<image:image>
<image:loc>https://mcxtywzsmeezbwapdahq.supabase.co/storage/v1/object/public/canvas-images/1ea5b494-0bbf-4b2b-b155-3eecc768ddf1/canvas/f21f0e25-0dd0-49e2-b6aa-ab4ef4acca9d/generated/13d08d04-d319-4dd6-8327-19fe12e077e8.jpeg</image:loc>
</image:image>
<lastmod>2026-10-28T12:21:36.640Z</lastmod>
<changefreq>monthly</changefreq>
<priority>0.7</priority>
</url>
<url>
<loc>https://inferenceengineering.co/posts/rdma-fundamentals-for-llm-infrastructure-engineers</loc>
<image:image>
<image:loc>https://mcxtywzsmeezbwapdahq.supabase.co/storage/v1/object/public/canvas-images/1ea5b494-0bbf-4b2b-b155-3eecc768ddf1/canvas/f21f0e25-0dd0-49e2-b6aa-ab4ef4acca9d/generated/5c28fcb5-8bd8-44a6-a37e-ea017eee18da.jpeg</image:loc>
</image:image>
<lastmod>2026-10-27T11:35:58.688Z</lastmod>
<changefreq>monthly</changefreq>
<priority>0.7</priority>
</url>
<url>
<loc>https://inferenceengineering.co/posts/kv-cache-tiering-across-dram-nvme-and-remote-storage-for-inference-latency</loc>
<image:image>
<image:loc>https://mcxtywzsmeezbwapdahq.supabase.co/storage/v1/object/public/canvas-images/1ea5b494-0bbf-4b2b-b155-3eecc768ddf1/canvas/f21f0e25-0dd0-49e2-b6aa-ab4ef4acca9d/generated/7c51ee0c-2eda-418f-8bfe-3b1ebf893ddf.jpeg</image:loc>
</image:image>
<lastmod>2026-10-26T12:06:36.955Z</lastmod>
<changefreq>monthly</changefreq>
<priority>0.7</priority>
</url>
<url>
<loc>https://inferenceengineering.co/posts/kv-cache-tiering-and-eviction-policy-design-for-128k-context-windows</loc>
<image:image>
<image:loc>https://mcxtywzsmeezbwapdahq.supabase.co/storage/v1/object/public/canvas-images/1ea5b494-0bbf-4b2b-b155-3eecc768ddf1/canvas/f21f0e25-0dd0-49e2-b6aa-ab4ef4acca9d/generated/49fa16d5-2a9c-4dc6-984d-2e957a392cd2.jpeg</image:loc>
</image:image>
<lastmod>2026-10-25T13:10:47.882Z</lastmod>
<changefreq>monthly</changefreq>
<priority>0.7</priority>
</url>
<url>
<loc>https://inferenceengineering.co/posts/kv-cache-lifecycle-orchestration-across-request-queues-and-preemption-boundaries</loc>
<image:image>
<image:loc>https://mcxtywzsmeezbwapdahq.supabase.co/storage/v1/object/public/canvas-images/1ea5b494-0bbf-4b2b-b155-3eecc768ddf1/canvas/f21f0e25-0dd0-49e2-b6aa-ab4ef4acca9d/generated/7332fc7e-8acb-4997-ad41-843c11e8d7c9.jpeg</image:loc>
</image:image>
<lastmod>2026-10-24T12:15:37.550Z</lastmod>
<changefreq>monthly</changefreq>
<priority>0.7</priority>
</url>
<url>
<loc>https://inferenceengineering.co/posts/object-storage-throughput-and-prefix-sharding-for-training-pipelines</loc>
<image:image>
<image:loc>https://mcxtywzsmeezbwapdahq.supabase.co/storage/v1/object/public/canvas-images/1ea5b494-0bbf-4b2b-b155-3eecc768ddf1/canvas/f21f0e25-0dd0-49e2-b6aa-ab4ef4acca9d/generated/77e0de6c-37ce-4bfc-935b-b9a343e4a4e4.jpeg</image:loc>
</image:image>
<lastmod>2026-10-23T13:21:41.805Z</lastmod>
<changefreq>monthly</changefreq>
<priority>0.7</priority>
</url>
<url>
<loc>https://inferenceengineering.co/posts/ssd-endurance-and-wear-management-in-kv-cache-offload-deployments</loc>
<image:image>
<image:loc>https://mcxtywzsmeezbwapdahq.supabase.co/storage/v1/object/public/canvas-images/1ea5b494-0bbf-4b2b-b155-3eecc768ddf1/canvas/f21f0e25-0dd0-49e2-b6aa-ab4ef4acca9d/generated/6e6c3dc3-0f38-4fe4-885e-2e94f5e45741.jpeg</image:loc>
</image:image>
<lastmod>2026-10-22T13:26:28.169Z</lastmod>
<changefreq>monthly</changefreq>
<priority>0.7</priority>
</url>
<url>
<loc>https://inferenceengineering.co/posts/gds-and-nvme-of-for-direct-storage-to-gpu-data-paths</loc>
<image:image>
<image:loc>https://mcxtywzsmeezbwapdahq.supabase.co/storage/v1/object/public/canvas-images/1ea5b494-0bbf-4b2b-b155-3eecc768ddf1/canvas/f21f0e25-0dd0-49e2-b6aa-ab4ef4acca9d/generated/faba2b52-a069-47e7-927d-2a7dae4d35aa.jpeg</image:loc>
</image:image>
<lastmod>2026-10-21T12:21:08.069Z</lastmod>
<changefreq>monthly</changefreq>
<priority>0.7</priority>
</url>
<url>
<loc>https://inferenceengineering.co/posts/filesystem-selection-for-high-throughput-ai-storage-nodes</loc>
<image:image>
<image:loc>https://mcxtywzsmeezbwapdahq.supabase.co/storage/v1/object/public/canvas-images/1ea5b494-0bbf-4b2b-b155-3eecc768ddf1/canvas/f21f0e25-0dd0-49e2-b6aa-ab4ef4acca9d/generated/88258302-9d2e-409f-a63f-1d399df9cc04.jpeg</image:loc>
</image:image>
<lastmod>2026-10-20T11:35:55.696Z</lastmod>
<changefreq>monthly</changefreq>
<priority>0.7</priority>
</url>
<url>
<loc>https://inferenceengineering.co/posts/storage-bottlenecks-during-distributed-training-checkpoint-saves</loc>
<image:image>
<image:loc>https://mcxtywzsmeezbwapdahq.supabase.co/storage/v1/object/public/canvas-images/1ea5b494-0bbf-4b2b-b155-3eecc768ddf1/canvas/f21f0e25-0dd0-49e2-b6aa-ab4ef4acca9d/generated/ac1f8c27-743c-47cd-a787-4072ea325341.jpeg</image:loc>
</image:image>
<lastmod>2026-10-19T12:06:06.886Z</lastmod>
<changefreq>monthly</changefreq>
<priority>0.7</priority>
</url>
<url>
<loc>https://inferenceengineering.co/posts/nfs-performance-for-ai-inference-model-loading</loc>
<image:image>
<image:loc>https://mcxtywzsmeezbwapdahq.supabase.co/storage/v1/object/public/canvas-images/1ea5b494-0bbf-4b2b-b155-3eecc768ddf1/canvas/f21f0e25-0dd0-49e2-b6aa-ab4ef4acca9d/generated/11534f8d-bf7d-4d8c-914b-2dae8e2a748a.jpeg</image:loc>
</image:image>
<lastmod>2026-10-18T13:05:44.586Z</lastmod>
<changefreq>monthly</changefreq>
<priority>0.7</priority>
</url>
<url>
<loc>https://inferenceengineering.co/posts/all-flash-storage-architecture-for-llm-training-data</loc>
<image:image>
<image:loc>https://mcxtywzsmeezbwapdahq.supabase.co/storage/v1/object/public/canvas-images/1ea5b494-0bbf-4b2b-b155-3eecc768ddf1/canvas/f21f0e25-0dd0-49e2-b6aa-ab4ef4acca9d/generated/f7673a55-ab52-4115-93d1-9c78a6945e12.jpeg</image:loc>
</image:image>
<lastmod>2026-10-17T12:45:29.997Z</lastmod>
<changefreq>monthly</changefreq>
<priority>0.7</priority>
</url>
<url>
<loc>https://inferenceengineering.co/posts/nvme-queue-depth-tuning-for-ai-checkpoint-workloads</loc>
<image:image>
<image:loc>https://mcxtywzsmeezbwapdahq.supabase.co/storage/v1/object/public/canvas-images/1ea5b494-0bbf-4b2b-b155-3eecc768ddf1/canvas/f21f0e25-0dd0-49e2-b6aa-ab4ef4acca9d/generated/1a257aa2-fdf2-49f0-bbf5-4e94dfbc2104.jpeg</image:loc>
</image:image>
<lastmod>2026-10-16T13:36:52.212Z</lastmod>
<changefreq>monthly</changefreq>
<priority>0.7</priority>
</url>
<url>
<loc>https://inferenceengineering.co/posts/unified-memory-and-managed-memory-pitfalls-in-cuda-inference</loc>
<image:image>
<image:loc>https://mcxtywzsmeezbwapdahq.supabase.co/storage/v1/object/public/canvas-images/1ea5b494-0bbf-4b2b-b155-3eecc768ddf1/canvas/f21f0e25-0dd0-49e2-b6aa-ab4ef4acca9d/generated/194fa19f-9790-4a3f-83eb-2e59d3a7cf68.jpeg</image:loc>
</image:image>
<lastmod>2026-10-14T12:46:16.624Z</lastmod>
<changefreq>monthly</changefreq>
<priority>0.7</priority>
</url>
<url>
<loc>https://inferenceengineering.co/posts/nvlink-vs-pcie-bandwidth-impact-on-kv-cache-transfers</loc>
<image:image>
<image:loc>https://mcxtywzsmeezbwapdahq.supabase.co/storage/v1/object/public/canvas-images/1ea5b494-0bbf-4b2b-b155-3eecc768ddf1/canvas/f21f0e25-0dd0-49e2-b6aa-ab4ef4acca9d/generated/668db018-3926-4e29-b08a-75379e2572a8.jpeg</image:loc>
</image:image>
<lastmod>2026-10-13T11:31:01.576Z</lastmod>
<changefreq>monthly</changefreq>
<priority>0.7</priority>
</url>
<url>
<loc>https://inferenceengineering.co/posts/data-residency-strategies-for-multi-tenant-inference</loc>
<image:image>
<image:loc>https://mcxtywzsmeezbwapdahq.supabase.co/storage/v1/object/public/canvas-images/1ea5b494-0bbf-4b2b-b155-3eecc768ddf1/canvas/f21f0e25-0dd0-49e2-b6aa-ab4ef4acca9d/generated/d616c983-b608-4560-b733-5da0c924cf78.jpeg</image:loc>
</image:image>
<lastmod>2026-10-12T12:31:53.954Z</lastmod>
<changefreq>monthly</changefreq>
<priority>0.7</priority>
</url>
<url>
<loc>https://inferenceengineering.co/posts/multi-instance-gpu-partitioning-for-inference-density</loc>
<image:image>
<image:loc>https://mcxtywzsmeezbwapdahq.supabase.co/storage/v1/object/public/canvas-images/1ea5b494-0bbf-4b2b-b155-3eecc768ddf1/canvas/f21f0e25-0dd0-49e2-b6aa-ab4ef4acca9d/generated/2fa2a7c1-94c8-4314-ad02-bf019015ce3b.jpeg</image:loc>
</image:image>
<lastmod>2026-10-10T13:20:40.911Z</lastmod>
<changefreq>monthly</changefreq>
<priority>0.7</priority>
</url>
<url>
<loc>https://inferenceengineering.co/posts/weight-loading-latency-and-cold-start-optimization</loc>
<image:image>
<image:loc>https://mcxtywzsmeezbwapdahq.supabase.co/storage/v1/object/public/canvas-images/1ea5b494-0bbf-4b2b-b155-3eecc768ddf1/canvas/f21f0e25-0dd0-49e2-b6aa-ab4ef4acca9d/generated/5676609c-bca0-45df-b9f6-cef8a22453f6.png</image:loc>
</image:image>
<lastmod>2026-10-09T11:50:37.701Z</lastmod>
<changefreq>monthly</changefreq>
<priority>0.7</priority>
</url>
<url>
<loc>https://inferenceengineering.co/posts/hbm-bandwidth-saturation-in-autoregressive-decoding</loc>
<image:image>
<image:loc>https://mcxtywzsmeezbwapdahq.supabase.co/storage/v1/object/public/canvas-images/1ea5b494-0bbf-4b2b-b155-3eecc768ddf1/canvas/f21f0e25-0dd0-49e2-b6aa-ab4ef4acca9d/generated/321225b0-7c33-4278-91ad-50941e4cb57f.jpeg</image:loc>
</image:image>
<lastmod>2026-10-08T11:30:33.901Z</lastmod>
<changefreq>monthly</changefreq>
<priority>0.7</priority>
</url>
<url>
<loc>https://inferenceengineering.co/posts/gpu-memory-hierarchy-for-llm-practitioners</loc>
<image:image>
<image:loc>https://mcxtywzsmeezbwapdahq.supabase.co/storage/v1/object/public/canvas-images/1ea5b494-0bbf-4b2b-b155-3eecc768ddf1/canvas/f21f0e25-0dd0-49e2-b6aa-ab4ef4acca9d/generated/e29e5c6d-7ba4-4cc3-bdc9-0ec5ac72c8ab.jpeg</image:loc>
</image:image>
<lastmod>2026-10-07T13:00:46.925Z</lastmod>
<changefreq>monthly</changefreq>
<priority>0.7</priority>
</url>
<url>
<loc>https://inferenceengineering.co/posts/chunked-prefill-and-its-effect-on-time-to-first-token</loc>
<image:image>
<image:loc>https://mcxtywzsmeezbwapdahq.supabase.co/storage/v1/object/public/canvas-images/1ea5b494-0bbf-4b2b-b155-3eecc768ddf1/canvas/f21f0e25-0dd0-49e2-b6aa-ab4ef4acca9d/generated/bacc6a3f-e4cd-4132-932b-c5ecc5cba2a3.png</image:loc>
</image:image>
<lastmod>2026-10-06T11:30:17.340Z</lastmod>
<changefreq>monthly</changefreq>
<priority>0.7</priority>
</url>
<url>
<loc>https://inferenceengineering.co/posts/scheduler-design-for-long-context-and-short-context-mixed-traffic</loc>
<image:image>
<image:loc>https://mcxtywzsmeezbwapdahq.supabase.co/storage/v1/object/public/canvas-images/1ea5b494-0bbf-4b2b-b155-3eecc768ddf1/canvas/f21f0e25-0dd0-49e2-b6aa-ab4ef4acca9d/generated/944c72a1-e32d-4762-b274-32e567dea73f.jpeg</image:loc>
</image:image>
<lastmod>2026-10-05T12:40:24.362Z</lastmod>
<changefreq>monthly</changefreq>
<priority>0.7</priority>
</url>
<url>
<loc>https://inferenceengineering.co/posts/continuous-batching-mechanics-and-head-of-line-blocking</loc>
<image:image>
<image:loc>https://mcxtywzsmeezbwapdahq.supabase.co/storage/v1/object/public/canvas-images/1ea5b494-0bbf-4b2b-b155-3eecc768ddf1/canvas/f21f0e25-0dd0-49e2-b6aa-ab4ef4acca9d/generated/41d8b706-f1a1-4c8c-bbee-6be1fce3ab27.jpeg</image:loc>
</image:image>
<lastmod>2026-10-04T11:39:32.227Z</lastmod>
<changefreq>monthly</changefreq>
<priority>0.7</priority>
</url>
<url>
<loc>https://inferenceengineering.co/posts/gpudirect-storage-integration-in-inference-pipelines</loc>
<image:image>
<image:loc>https://mcxtywzsmeezbwapdahq.supabase.co/storage/v1/object/public/canvas-images/1ea5b494-0bbf-4b2b-b155-3eecc768ddf1/canvas/f21f0e25-0dd0-49e2-b6aa-ab4ef4acca9d/generated/8da27df6-0af9-4be6-b9ab-cc58c18f9314.jpeg</image:loc>
</image:image>
<lastmod>2026-10-03T12:54:56.782Z</lastmod>
<changefreq>monthly</changefreq>
<priority>0.7</priority>
</url>
<url>
<loc>https://inferenceengineering.co/posts/inference-on-open-source-models-vs-proprietary-apis-at-scale</loc>
<image:image>
<image:loc>https://mcxtywzsmeezbwapdahq.supabase.co/storage/v1/object/public/canvas-images/1ea5b494-0bbf-4b2b-b155-3eecc768ddf1/canvas/f21f0e25-0dd0-49e2-b6aa-ab4ef4acca9d/generated/e4a6f3d8-548e-4190-8c82-781eba70a874.jpeg</image:loc>
</image:image>
<lastmod>2026-10-02T11:35:03.818Z</lastmod>
<changefreq>monthly</changefreq>
<priority>0.7</priority>
</url>
<url>
<loc>https://inferenceengineering.co/posts/tensor-parallelism-and-kv-cache-sharding-patterns</loc>
<image:image>
<image:loc>https://mcxtywzsmeezbwapdahq.supabase.co/storage/v1/object/public/canvas-images/1ea5b494-0bbf-4b2b-b155-3eecc768ddf1/canvas/f21f0e25-0dd0-49e2-b6aa-ab4ef4acca9d/generated/37db8d6b-9b93-42e1-81e6-2888da190e10.jpeg</image:loc>
</image:image>
<lastmod>2026-10-01T13:25:37.695Z</lastmod>
<changefreq>monthly</changefreq>
<priority>0.7</priority>
</url>
<url>
<loc>https://inferenceengineering.co/posts/speculative-decoding-storage-and-memory-overhead</loc>
<image:image>
<image:loc>https://mcxtywzsmeezbwapdahq.supabase.co/storage/v1/object/public/canvas-images/1ea5b494-0bbf-4b2b-b155-3eecc768ddf1/canvas/f21f0e25-0dd0-49e2-b6aa-ab4ef4acca9d/generated/e4e6b650-4784-4123-a2b1-3e195d5c0dc1.jpeg</image:loc>
</image:image>
<lastmod>2026-09-30T17:36:22.121Z</lastmod>
<changefreq>monthly</changefreq>
<priority>0.7</priority>
</url>
<url>
<loc>https://inferenceengineering.co/posts/kv-cache-sizing-for-concurrent-request-budgets</loc>
<image:image>
<image:loc>https://mcxtywzsmeezbwapdahq.supabase.co/storage/v1/object/public/canvas-images/1ea5b494-0bbf-4b2b-b155-3eecc768ddf1/canvas/f21f0e25-0dd0-49e2-b6aa-ab4ef4acca9d/generated/775c9878-ff50-4e3a-8fb4-bc97991a285c.jpeg</image:loc>
</image:image>
<lastmod>2026-09-30T02:25:44.078Z</lastmod>
<changefreq>monthly</changefreq>
<priority>0.7</priority>
</url>
<url>
<loc>https://inferenceengineering.co/posts/distributed-kv-cache-across-inference-nodes</loc>
<image:image>
<image:loc>https://mcxtywzsmeezbwapdahq.supabase.co/storage/v1/object/public/canvas-images/1ea5b494-0bbf-4b2b-b155-3eecc768ddf1/canvas/f21f0e25-0dd0-49e2-b6aa-ab4ef4acca9d/generated/3c08d3d4-0feb-4547-9340-710c89568c4a.png</image:loc>
</image:image>
<lastmod>2026-09-26T23:04:28.713Z</lastmod>
<changefreq>monthly</changefreq>
<priority>0.7</priority>
</url>
<url>
<loc>https://inferenceengineering.co/posts/kv-cache-offload-to-nvme-under-memory-pressure</loc>
<image:image>
<image:loc>https://mcxtywzsmeezbwapdahq.supabase.co/storage/v1/object/public/canvas-images/1ea5b494-0bbf-4b2b-b155-3eecc768ddf1/canvas/f21f0e25-0dd0-49e2-b6aa-ab4ef4acca9d/generated/05a370fa-d54e-43db-84e3-14ea592e0cab.jpeg</image:loc>
</image:image>
<lastmod>2026-09-26T04:08:00.736Z</lastmod>
<changefreq>monthly</changefreq>
<priority>0.7</priority>
</url>
<url>
<loc>https://inferenceengineering.co/posts/prefix-caching-and-prompt-deduplication-in-vllm</loc>
<image:image>
<image:loc>https://mcxtywzsmeezbwapdahq.supabase.co/storage/v1/object/public/canvas-images/1ea5b494-0bbf-4b2b-b155-3eecc768ddf1/canvas/f21f0e25-0dd0-49e2-b6aa-ab4ef4acca9d/generated/d4ce559d-b5e0-4d83-ad5e-7fbc3cf89e8e.jpeg</image:loc>
</image:image>
<lastmod>2026-09-24T11:19:19.335Z</lastmod>
<changefreq>monthly</changefreq>
<priority>0.7</priority>
</url>
<url>
<loc>https://inferenceengineering.co/posts/paged-attention-and-kv-cache-fragmentation</loc>
<image:image>
<image:loc>https://mcxtywzsmeezbwapdahq.supabase.co/storage/v1/object/public/canvas-images/1ea5b494-0bbf-4b2b-b155-3eecc768ddf1/canvas/f21f0e25-0dd0-49e2-b6aa-ab4ef4acca9d/generated/0e55ff8b-3605-4d84-bfa0-608abfa06d18.jpeg</image:loc>
</image:image>
<lastmod>2026-09-22T11:48:05.443Z</lastmod>
<changefreq>monthly</changefreq>
<priority>0.7</priority>
</url>
<url>
<loc>https://inferenceengineering.co/posts/kv-cache-memory-layout-on-multi-gpu-hosts</loc>
<image:image>
<image:loc>https://mcxtywzsmeezbwapdahq.supabase.co/storage/v1/object/public/canvas-images/1ea5b494-0bbf-4b2b-b155-3eecc768ddf1/canvas/f21f0e25-0dd0-49e2-b6aa-ab4ef4acca9d/generated/6a89ac59-1c25-473a-b25c-e4c95a7c9d87.jpeg</image:loc>
</image:image>
<lastmod>2026-09-21T12:49:16.366Z</lastmod>
<changefreq>monthly</changefreq>
<priority>0.7</priority>
</url>
<url>
<loc>https://inferenceengineering.co/posts/lru-vs-gdsf-vs-recency-weighted-eviction-for-long-context-kv-cache</loc>
<image:image>
<image:loc>https://mcxtywzsmeezbwapdahq.supabase.co/storage/v1/object/public/canvas-images/1ea5b494-0bbf-4b2b-b155-3eecc768ddf1/canvas/f21f0e25-0dd0-49e2-b6aa-ab4ef4acca9d/generated/219447c9-a215-4cb7-b148-5166c56ae5dd.jpeg</image:loc>
</image:image>
<lastmod>2026-09-18T12:58:36.172Z</lastmod>
<changefreq>monthly</changefreq>
<priority>0.7</priority>
</url>
<url>
<loc>https://inferenceengineering.co/posts/prefill-and-decode-disaggregation-in-llm-serving</loc>
<image:image>
<image:loc>https://mcxtywzsmeezbwapdahq.supabase.co/storage/v1/object/public/canvas-images/1ea5b494-0bbf-4b2b-b155-3eecc768ddf1/canvas/f21f0e25-0dd0-49e2-b6aa-ab4ef4acca9d/generated/572be82f-2b4c-4f93-a481-e198d658563d.jpeg</image:loc>
</image:image>
<lastmod>2026-09-18T06:51:30.742Z</lastmod>
<changefreq>monthly</changefreq>
<priority>0.7</priority>
</url>
</urlset>
