https://tarundecipher.com 2026-08-01T22:59:26.860Z https://tarundecipher.com/articles/how-openai-scaled-postgresql-and-what-to-do-in-your-system 2026-06-29T16:29:29.347Z https://tarundecipher.com/articles/how-flink-works-internally 2026-06-29T16:29:29.309Z https://tarundecipher.com/articles/why-netflix-4k-can-look-worse-than-theater-1080p 2026-06-29T16:29:29.274Z https://tarundecipher.com/articles/how-to-deduplicate-streaming-events-at-scale 2026-06-29T16:29:29.238Z https://tarundecipher.com/articles/why-transformers-use-residual-connections 2026-06-29T16:29:29.207Z https://tarundecipher.com/articles/what-is-a-feed-forward-network-in-a-transformer 2026-06-29T16:29:29.169Z https://tarundecipher.com/articles/why-multiple-attention-heads-help 2026-06-29T16:29:29.127Z https://tarundecipher.com/articles/what-is-multi-head-attention 2026-06-29T16:29:29.087Z https://tarundecipher.com/articles/why-attention-uses-q-k-and-v 2026-06-29T16:29:29.045Z https://tarundecipher.com/articles/what-are-query-key-and-value-vectors 2026-06-29T16:29:29.002Z https://tarundecipher.com/articles/why-transformers-need-positional-information 2026-06-29T16:29:28.958Z https://tarundecipher.com/articles/what-are-embeddings 2026-06-29T16:29:28.912Z https://tarundecipher.com/articles/why-transformers-replaced-rnns-for-language-modeling 2026-06-29T16:29:28.874Z https://tarundecipher.com/articles/what-is-a-transformer 2026-06-29T16:29:28.841Z https://tarundecipher.com/articles/why-self-attention-can-look-at-every-token 2026-06-29T16:29:28.774Z https://tarundecipher.com/articles/what-is-self-attention 2026-06-29T16:29:28.739Z https://tarundecipher.com/articles/what-problem-did-attention-solve 2026-06-29T16:29:28.705Z https://tarundecipher.com/articles/why-sequence-models-struggled-with-long-context 2026-06-29T16:29:28.659Z https://tarundecipher.com/articles/why-lstms-were-better-than-plain-rnns 2026-06-29T16:29:28.609Z https://tarundecipher.com/articles/why-rnns-process-text-sequentially 2026-06-29T16:29:28.560Z https://tarundecipher.com/articles/why-n-gram-models-hit-a-wall 2026-06-29T16:29:28.518Z https://tarundecipher.com/articles/what-came-before-transformers 2026-06-29T16:29:28.483Z https://tarundecipher.com/articles/what-problem-did-language-models-try-to-solve 2026-06-29T16:29:28.446Z https://tarundecipher.com/articles/how-speculative-decoding-makes-llms-feel-faster 2026-06-29T16:29:28.404Z https://tarundecipher.com/articles/why-kafka-needed-zookeeper 2026-06-29T16:29:28.363Z https://tarundecipher.com/articles/why-columnar-databases-feel-like-cheating 2026-06-29T16:29:28.324Z https://tarundecipher.com/articles/llm-burners-move-models-into-silicon 2026-06-29T16:29:28.251Z https://tarundecipher.com/articles/fast-llm-serving-is-memory-management 2026-06-29T16:29:28.212Z https://tarundecipher.com/articles/context-length-is-gpu-memory-too 2026-06-27T14:37:38.760Z https://tarundecipher.com/topics/ai 2026-08-01T22:59:26.860Z https://tarundecipher.com/topics/data-engineering 2026-08-01T22:59:26.860Z https://tarundecipher.com/paths 2026-08-01T22:59:26.860Z https://tarundecipher.com/paths/llm-inference-systems 2026-08-01T22:59:26.860Z https://tarundecipher.com/paths/serving-performance 2026-08-01T22:59:26.860Z https://tarundecipher.com/paths/data-systems-internals 2026-08-01T22:59:26.860Z https://tarundecipher.com/issues/llm-inference-notes 2026-06-27T14:37:38.760Z