--- title: "#llm-architecture articles | Hashnode" description: "Explore 15 articles tagged #llm-architecture on Hashnode. Tutorials, guides, and how-tos from the community." url: "https://hashnode.com/tag/llm-architecture" --- Tag # #llm-architecture 15 posts · 0 followers [Ignacio Di Rocco](https://hashnode.com/@idirocco) · [Oct 28](https://hashnode.com/posts/a-small-pipeline-with-explicit-boundaries-architecture-implementation/6ae1d3e7a82de6eb4e6a2671) ### [A Small Pipeline, With Explicit Boundaries: Architecture & Implementation](https://idirocco.hashnode.dev/a-small-pipeline-with-explicit-boundaries-architecture-implementation?utm_source=hashnode&utm_medium=feed) Model choices, orchestration, structured outputs, evaluation, and the production trade-offs I haven't measured yet. 6 min read · [idirocco.hashnode.dev](https://idirocco.hashnode.dev/a-small-pipeline-with-explicit-boundaries-architecture-implementation?utm_source=hashnode&utm_medium=feed) [Rahul Sai Indeevar V](https://hashnode.com/@rahul-ai) · [Sep 5](https://hashnode.com/posts/llm-architectures-from-bert-and-gpt-2-to-xlnet-and-t5/6a9c53131167f83341f0de4b) ### [LLM Architectures: From BERT and GPT-2 to XLNet and T5](https://rahul-ai.hashnode.dev/llm-architectures-from-bert-and-gpt-2-to-xlnet-and-t5?utm_source=hashnode&utm_medium=feed) Understanding the internal mechanics of Large Language Model (LLM) architectures is crucial for building, fine-tuning, or deploying modern AI models. Why does GPT excel at text generation, while BERT 9 min read · [rahul-ai.hashnode.dev](https://rahul-ai.hashnode.dev/llm-architectures-from-bert-and-gpt-2-to-xlnet-and-t5?utm_source=hashnode&utm_medium=feed) [Helkyn Coello](https://hashnode.com/@helkyn-coello) · [Sep 5](https://hashnode.com/posts/financial-analysis-needs-ai-at-the-edges-never-in-the-math/6a9bfe7dd348da384fbf9f66) ### [Financial Analysis Needs AI at the Edges. Never in the Math.](https://unlockedconsulting.hashnode.dev/financial-analysis-needs-ai-at-the-edges-never-in-the-math?utm_source=hashnode&utm_medium=feed) A 33-metric accounting pipeline where AI parses PDFs at the entrance and writes summary at the exit, while deterministic code computes every number 8 min read · [unlockedconsulting.hashnode.dev](https://unlockedconsulting.hashnode.dev/financial-analysis-needs-ai-at-the-edges-never-in-the-math?utm_source=hashnode&utm_medium=feed) [Rajshree](https://hashnode.com/@rjshree) · [Aug 22](https://hashnode.com/posts/your-rag-isn-t-broken-your-retrieval-pipeline-is/6a89c7a4ca64fd0af274c690) ### [Your RAG Isn't Broken. Your Retrieval Pipeline Is.](https://rjshree.hashnode.dev/your-rag-isn-t-broken-your-retrieval-pipeline-is?utm_source=hashnode&utm_medium=feed) A practical deep dive into why RAG systems fail. Learn how chunking, embeddings, retrieval, metadata, reranking, query rewriting, context quality, and evaluation determine RAG accuracy." excerpt: "Most RAG failures aren't caused by the LLM. They begin much earlier—inside the retrieval pipeline. Learn how to diagnose and fix the real problems behind irrelevant context, hallucinations, poor answers, and unreliable RAG systems." slug: "your-rag-isnt-broken-your-retrieval-pipeline-is 25 min read · [rjshree.hashnode.dev](https://rjshree.hashnode.dev/your-rag-isn-t-broken-your-retrieval-pipeline-is?utm_source=hashnode&utm_medium=feed) [Learn with HJ](https://hashnode.com/@hardeepjethwani) · [Jul 11](https://hashnode.com/posts/mixture-of-experts-why-some-models-use-specialist-sub-brains/6a5262b66aa432b551ce7b7e) ### [Mixture of Experts: Why Some Models Use Specialist Sub-Brains](https://hardeepjethwani.hashnode.dev/mixture-of-experts-why-some-models-use-specialist-sub-brains?utm_source=hashnode&utm_medium=feed) Explain MoE architecture simply, benefits, routing, cost, and limitations. 6 min read · [hardeepjethwani.hashnode.dev](https://hardeepjethwani.hashnode.dev/mixture-of-experts-why-some-models-use-specialist-sub-brains?utm_source=hashnode&utm_medium=feed) [Ujjwal Tripathi](https://hashnode.com/@ujjwalt) · [Jul 1](https://hashnode.com/posts/how-to-build-an-ai-agent-from-scratch-in-2025-architecture-stack/6a4507591275adf1eb2c3dca) ### [How to Build an AI Agent from Scratch in 2025: Architecture & Stack](https://microcosmworks.hashnode.dev/how-to-build-an-ai-agent-from-scratch-in-2025-architecture-stack?utm_source=hashnode&utm_medium=feed) A practical engineering guide to building production-ready AI agents in 2025 — covering ReAct, memory design, tool layers, stack decisions, and the failure modes most teams hit after demo day. 6 min read · [microcosmworks.hashnode.dev](https://microcosmworks.hashnode.dev/how-to-build-an-ai-agent-from-scratch-in-2025-architecture-stack?utm_source=hashnode&utm_medium=feed) [Nishil Bhave](https://hashnode.com/@nishilbhave) · [May 4](https://hashnode.com/posts/ai-gateway-architecture-7-cross-cutting-concerns-2026/69f89f1aa6f8d74aeb212870) ### [AI Gateway Architecture: 7 Cross-Cutting Concerns (2026)](https://nishilbhave.hashnode.dev/ai-gateway-architecture-7-cross-cutting-concerns-2026?utm_source=hashnode&utm_medium=feed) Inside an AI Gateway: 7 Cross-Cutting Concerns That Don't Belong in Your App Code Pull up the source of any production AI feature shipped in 2026 and you'll find the same mess. An OpenAI client wrapped in a try/except. A hardcoded retry. An if model ... 17 min read · [nishilbhave.hashnode.dev](https://nishilbhave.hashnode.dev/ai-gateway-architecture-7-cross-cutting-concerns-2026?utm_source=hashnode&utm_medium=feed) [Anup Karanjkar](https://hashnode.com/@anupkaranjkar) · [Apr 27](https://hashnode.com/posts/claude-mythos-5-10-trillion-parameter-moe-developer-guide-2026/69efc0473cfb0e9de43df478) ### [Claude Mythos 5: Anthropic's 10-Trillion Parameter MoE Model — Complete Developer Guide (April 2026)](https://wowhow.hashnode.dev/claude-mythos-5-10-trillion-parameter-moe-developer-guide-2026?utm_source=hashnode&utm_medium=feed) Claude Mythos 5 uses 10T parameters (MoE, ~1T active) to set a new frontier. Architecture, $40-60/M pricing, Glasswing access, and a decision framework inside. 10 min read · [wowhow.hashnode.dev](https://wowhow.hashnode.dev/claude-mythos-5-10-trillion-parameter-moe-developer-guide-2026?utm_source=hashnode&utm_medium=feed) [O Meling](https://hashnode.com/@gothar) · [Apr 26](https://hashnode.com/posts/the-librarian-principle-why-your-rag-system-doesnt-need-a-vector-database/69edc1ed3273eb5359c9c5ec) ### [The Librarian Principle: Why Your RAG System Doesn't Need a Vector Database](https://gothartech.hashnode.dev/the-librarian-principle-why-your-rag-system-doesnt-need-a-vector-database?utm_source=hashnode&utm_medium=feed) Most RAG architectures bury their smartest component under layers of plumbing. We built a production retrieval system with SQLite, five MCP tools, and zero embeddings. It won. 8 min read · [gothartech.hashnode.dev](https://gothartech.hashnode.dev/the-librarian-principle-why-your-rag-system-doesnt-need-a-vector-database?utm_source=hashnode&utm_medium=feed) [Jangwook Kim](https://hashnode.com/@effloow) · [Apr 18](https://hashnode.com/posts/ai-context-window-race-1m-tokens-developer-guide-2026/69e33ffe1caaf0723ee4cbea) ### [The AI Context Window Race: What 1M Tokens Means for Devs](https://effloow.hashnode.dev/ai-context-window-race-1m-tokens-developer-guide-2026?utm_source=hashnode&utm_medium=feed) The headline says "1 million tokens." The fine print says something different. In April 2026, five frontier models sit at or above the 1 million token mark. One — Meta's Llama 4 Scout — claims 10 million. Numbers like these appear in every vendor ann... 11 min read · [effloow.hashnode.dev](https://effloow.hashnode.dev/ai-context-window-race-1m-tokens-developer-guide-2026?utm_source=hashnode&utm_medium=feed)