Applied AI & LLMs
These articles treat AI as part of the architecture, not as a magic dependency. The Inside LLMs series explains what actually happens between a web crawl and a chat assistant, so the trade-offs later on make sense.
The production articles cover the other half: running models on Amazon Bedrock, picking the right model per message to control cost, and keeping orchestration small enough to debug.
Start here
-
From Common Crawl to the Base Model: How an LLM Learns Language
How an LLM goes from raw web data to the base model: the Transformer, Common Crawl, FineWeb, tokenization and pre-training explained clearly, no hype.
-
Amazon Bedrock in practice: AI as part of the architecture, not as an external dependency
How I moved from OpenAI to Amazon Bedrock: IAM auth, consolidated billing, Haiku vs Sonnet by task, and real trade-offs of running AI on AWS.
-
Decision Models in Practice: When Picking the LLM per Message Pays Off (and When the Mid-Tier Model Is Enough)
I had Microsoft-Decision-1 pick the subagent and LLM in a post-sales chatbot: harness, calibration, cost guard, and what 1,000 synthetic conversations showed.
Series in this guide
All articles (6)
-
LangGraph Only for the Handoff: A Support Chatbot Where a Decision Model Picks Who Answers
A post-sales chatbot built with LangChain and LangGraph where a decision model picks the subagent per message: architecture, code, fallback, A/B and failures.
-
Decision Models in Practice: When Picking the LLM per Message Pays Off (and When the Mid-Tier Model Is Enough)
I had Microsoft-Decision-1 pick the subagent and LLM in a post-sales chatbot: harness, calibration, cost guard, and what 1,000 synthetic conversations showed.
-
From Base Model to ChatGPT: SFT, Tools and Reinforcement Learning
How an LLM base model becomes ChatGPT: supervised fine-tuning, tool use, hallucination reduction and reinforcement learning explained clearly and in depth.
-
From Common Crawl to the Base Model: How an LLM Learns Language
How an LLM goes from raw web data to the base model: the Transformer, Common Crawl, FineWeb, tokenization and pre-training explained clearly, no hype.
-
Amazon Bedrock in practice: AI as part of the architecture, not as an external dependency
How I moved from OpenAI to Amazon Bedrock: IAM auth, consolidated billing, Haiku vs Sonnet by task, and real trade-offs of running AI on AWS.
-
What I have learned from building a product with Spec-Driven Development
What I learned building a product with Spec-Driven Development: How structured specs eliminate technical debt using AI as a development peer.