--- title: "Sriram Krishnan (@sriramkrishnan) | Hashnode" description: "Data | BI Solutions Architect | AI. Read the latest articles by Sriram Krishnan on Hashnode." url: "https://hashnode.com/@sriramkrishnan" --- @sriramkrishnan # Sriram Krishnan @sriramkrishnan · Joined June 2025 Data | BI Solutions Architect | AI ### [LLM Deep Dive — Part 2 Post Training](https://moderndataarchitect.hashnode.dev/llm-deep-dive-part-2-post-training?utm_source=hashnode&utm_medium=feed) This is a series based on Andrej Karpathy’s “Deep Dive into LLMs like ChatGPT.” Part 1 of the series is here. A base model can be thought of as a token-level simulator of the internet. At its core, it does not reason or plan — it simply predicts the ... 8 min read · [moderndataarchitect.hashnode.dev](https://moderndataarchitect.hashnode.dev/llm-deep-dive-part-2-post-training?utm_source=hashnode&utm_medium=feed) [Dec 29, 2025](https://hashnode.com/posts/llm-deep-dive-part-2-post-training/6952622c4a069ae694cf6b73) ### [LLM Deep Dive — Part 1 Pretraining](https://moderndataarchitect.hashnode.dev/llm-deep-dive-part-1-pretraining?utm_source=hashnode&utm_medium=feed) This is a series based on Andrej Karpathy’s “Deep Dive into LLMs like ChatGPT.” The goal of this series is to peel back the layers and understand what actually happens under the hood when a Large Language Model (LLM) is built. LLMs often feel intelli... 8 min read · [moderndataarchitect.hashnode.dev](https://moderndataarchitect.hashnode.dev/llm-deep-dive-part-1-pretraining?utm_source=hashnode&utm_medium=feed) [Dec 27, 2025](https://hashnode.com/posts/llm-deep-dive-part-1-pretraining/694fbb48c0b285895296c5a3) ### [A Design-First Approach to Conceptual Data Modelling](https://moderndataarchitect.hashnode.dev/a-design-first-approach-to-conceptual-data-modelling?utm_source=hashnode&utm_medium=feed) Many complex data projects stumble before they even begin. Teams rush to select technology, build data pipelines, and design dashboards, only to find themselves bogged down in rework, misalignment, and a final product that fails to meet business need... 5 min read · [moderndataarchitect.hashnode.dev](https://moderndataarchitect.hashnode.dev/a-design-first-approach-to-conceptual-data-modelling?utm_source=hashnode&utm_medium=feed) [Nov 10, 2025](https://hashnode.com/posts/a-design-first-approach-to-conceptual-data-modelling/69121354e4789e0b7413e9be) ### [The Architect's Blueprint – A 4-Week Data Architecture Sprint](https://moderndataarchitect.hashnode.dev/the-architects-blueprint-a-4-week-data-architecture-sprint?utm_source=hashnode&utm_medium=feed) Over this series, we have assembled all the powerful components of a modern data platform: a scalable cloud warehouse, a flexible Lakehouse, robust transformation frameworks, and an active semantic layer. We have the "what." But a collection of high-... 6 min read · [moderndataarchitect.hashnode.dev](https://moderndataarchitect.hashnode.dev/the-architects-blueprint-a-4-week-data-architecture-sprint?utm_source=hashnode&utm_medium=feed) [Oct 23, 2025](https://hashnode.com/posts/the-architects-blueprint-a-4-week-data-architecture-sprint/68f9cdd10fd5227b702cde35) ### [Metadata – From Passive Documentation to the Active Engine of AI](https://moderndataarchitect.hashnode.dev/metadata-from-passive-documentation-to-the-active-engine-of-ai?utm_source=hashnode&utm_medium=feed) For years, metadata has been treated like the dustiest book in the library. We knew it was important, but creating it was a chore—an afterthought to the "real work" of building pipelines. That era is over. The rise of self-service BI and, more import... 4 min read · [moderndataarchitect.hashnode.dev](https://moderndataarchitect.hashnode.dev/metadata-from-passive-documentation-to-the-active-engine-of-ai?utm_source=hashnode&utm_medium=feed) [Oct 21, 2025](https://hashnode.com/posts/metadata-from-passive-documentation-to-the-active-engine-of-ai/68f7319408a214358b9f4f91)