AI
Generative AI explained in plain words: LLMs, RAG, embeddings, knowledge graphs, evals and AI agents, with real examples and a working reference build.
Part 1: GenAI and Knowledge Modeling
Generative AI in Plain Words: What LLMs Can and Can't Do
Generative AI creates text, images and code by predicting the next word. Here's what that means, and why your company's knowledge is the part it's missing.
Taxonomy: Organize Knowledge So AI Searches Less and Finds More
A taxonomy is a tree of categories. Tagging documents with one lets an AI assistant skip irrelevant content before it searches, which means less noise and more accurate answers.
Ontology: The Blueprint Behind Enterprise Knowledge
A taxonomy says what is a type of what. An ontology also says how things relate and which rules they follow. Here's the difference, with Schema.org as a real-world example.
Knowledge Graphs: Answering the Questions Vector Search Can't
Knowledge graphs store facts as a network of relationships. They answer chain questions like "who approves Rahul's leave?" that text search struggles with, and they pair well with RAG.
SHACL: Validation Rules That Keep LLM-Extracted Data Honest
SHACL is a set of validation rules for knowledge graph data, like form validation for facts. It catches an LLM's extraction mistakes before they are stored.
Part 2: Retrieval (RAG)
RAG Explained: Give Your LLM an Open-Book Exam
Retrieval-Augmented Generation searches your own documents first, then has the LLM answer from what it found, with a citation. Here's how the pipeline works.
Chunking Strategies: How to Cut Documents Without Losing Meaning
Chunking splits big documents into searchable pieces. Cut in the wrong place and the answer loses half its meaning. Here are the five common methods and when to use each.
Embeddings: Google Maps Coordinates for Meaning
An embedding turns text into numbers that capture its meaning, so "sick days" and "medical leave" end up close together. This is what makes semantic search work.
Choosing an Embedding Model: Dimensions, Storage and Accuracy
More embedding dimensions mean better accuracy but more storage and slower search. Here's a comparison of popular models, what they cost at a million chunks, and three rules for choosing.
Vector Databases, Top-K and Hybrid Search
A vector database finds the chunks closest in meaning to a question. Keyword search finds exact terms. Hybrid search uses both, and usually beats either one alone.
Re-ranking: The Second Opinion That Picks the Best Chunks
A re-ranker reads the question and each candidate together and scores how relevant it really is. It's slow but accurate, so you use it on a shortlist. Plus a real lesson on where re-rankers fail.
Part 3: Evals and Accuracy
Golden Evals: The Answer Key Every AI Feature Needs
A golden eval is a fixed set of real questions with expert-approved answers. Re-run it after every change, keep what improves the score and reject what doesn't. Here's how to build one.
Measuring AI Accuracy: Retrieval Metrics vs Answer Metrics
Measure retrieval and answers separately, because the fix depends on which one failed. A plain guide to Recall@K, MRR, faithfulness, correctness and LLM-as-judge.
Which Model for Which Job: Picking LLMs by Task, Cost and Evals
Use the smallest, cheapest model that still passes your eval. A practical map of which model types fit which tasks, plus model routing.
Part 4: Agentic Automation
AI Agents: From Chatbots That Talk to Assistants That Act
A chatbot tells you what to do. An agent plans, calls tools, checks results and repeats until the job is done. Here's a real agent trace and the guardrails that keep it safe.
Tool Calling: How LLMs Use Your APIs Without Running Code
With tool calling, the LLM picks a tool from a menu and asks for it. Your code runs it and hands back the result. The model never touches your systems directly.
MCP (Model Context Protocol): The USB-C Cable for AI
MCP is an open standard for connecting AI apps to tools and data. Build a connector once and every MCP-compatible app can use it. Here's what an MCP server offers.
Letting AI Agents Use the Command Line, Safely
Agents can run ordinary terminal commands like git, kubectl and aws because LLMs already know them. That's fast and powerful, so lock the dangerous buttons. CLI vs MCP compared.
Have a product idea or a system to scale?
Tell us what you're building. You'll hear back within one business day - from a senior engineer, not a sales rep.
- Reply within 24h
- Senior engineers only
- Fixed-scope or T&M
- NDA on request