Prompt/Deploy

Prompt/Deploy

Engineering AI features you'd be on call for

I consent to receive newsletters via email. Terms of use and Privacy policy.

Connect

Featured Posts

production-ragproduction-rag
The Chunking Problem: Size, Overlap, and How to Measure What Works

The Chunking Problem: Size, Overlap, and How to Measure What Works

How to think through chunking strategy and build the eval loop that tells you if it's working.

Hou C.
Hou C.
production-ragproduction-rag
Incremental Re-indexing and the Embedding Pipeline Nobody Talks About

Incremental Re-indexing and the Embedding Pipeline Nobody Talks About

Embedding once is easy. Here's the production engineering behind keeping your RAG index correct, fresh, and survivable when you need to change your model or chunking strategy

Hou C.
Hou C.
production-ragproduction-rag
How I'd Choose a Vector Database

How I'd Choose a Vector Database

The tutorial picks one for you. Here's the decision logic for choosing a vector database in production — so you can eliminate the wrong options before they become expensive ones.

Hou C.
Hou C.
production-ragproduction-rag
When RAG Is the Wrong Answer (And What to Reach For)

When RAG Is the Wrong Answer (And What to Reach For)

You've already shipped RAG. Now some use cases are failing structurally. A four-gate decision tree for knowing when to fix the pipeline vs. when to reach for something else.

Hou C.
Hou C.
Mental Models for Production AIMental Models for Production AI
Three Questions I’d Want Answered Before Building Any AI Feature

Three Questions I’d Want Answered Before Building Any AI Feature

Here's the three-question framework I use to decide whether an AI feature is worth building.

Hou C.
Hou C.
Agentic Content Platforms for Technical EducationAgentic Content Platforms for Technical Education
System Design Notes: Agentic Content Platforms for Technical Education — Series Overview

System Design Notes: Agentic Content Platforms for Technical Education — Series Overview

Designing an agentic content platform for technical education from the ground up — starting with evaluation, not generation

Hou C.
Hou C.
Agentic Content Platforms for Technical EducationAgentic Content Platforms for Technical Education
Stateful Orchestration for Reliable Course Production

Stateful Orchestration for Reliable Course Production

Production agent pipelines fail silently. Event-driven orchestration with idempotency, failure isolation, and content versioning keeps course content consistent and recoverable.

Hou C.
Hou C.
Agentic Content Platforms for Technical EducationAgentic Content Platforms for Technical Education
The Multi-Agent Course Artifact Pipeline

The Multi-Agent Course Artifact Pipeline

Design a six-agent content pipeline that mirrors curriculum team roles, with explicit tool constraints, shared state contracts, and isolated evaluation for each stage.

Hou C.
Hou C.
Agentic Content Platforms for Technical EducationAgentic Content Platforms for Technical Education
Evaluation-First Agent Architecture for Learning Outcomes

Evaluation-First Agent Architecture for Learning Outcomes

A practical architecture for LLM evaluation systems with calibrated judges, threshold-based routing, and dimension-specific HITL.

Hou C.
Hou C.
Agentic Content Platforms for Technical EducationAgentic Content Platforms for Technical Education
Why Content Generation Is the Wrong Goal for Technical Education

Why Content Generation Is the Wrong Goal for Technical Education

Agentic content systems for technical education should measure learning outcomes—and that requires designing evaluation before content.

Hou C.
Hou C.

Archive

production-ragproduction-rag
The Chunking Problem: Size, Overlap, and How to Measure What Works

The Chunking Problem: Size, Overlap, and How to Measure What Works

How to think through chunking strategy and build the eval loop that tells you if it's working.

Hou C.
Hou C.
production-ragproduction-rag
Incremental Re-indexing and the Embedding Pipeline Nobody Talks About

Incremental Re-indexing and the Embedding Pipeline Nobody Talks About

Embedding once is easy. Here's the production engineering behind keeping your RAG index correct, fresh, and survivable when you need to change your model or chunking strategy

Hou C.
Hou C.
production-ragproduction-rag
How I'd Choose a Vector Database

How I'd Choose a Vector Database

The tutorial picks one for you. Here's the decision logic for choosing a vector database in production — so you can eliminate the wrong options before they become expensive ones.

Hou C.
Hou C.
production-ragproduction-rag
When RAG Is the Wrong Answer (And What to Reach For)

When RAG Is the Wrong Answer (And What to Reach For)

You've already shipped RAG. Now some use cases are failing structurally. A four-gate decision tree for knowing when to fix the pipeline vs. when to reach for something else.

Hou C.
Hou C.
Mental Models for Production AIMental Models for Production AI
When Should I Retrain or Tune?

When Should I Retrain or Tune?

The decision framework for updates. An interactive guide to knowing when to tune prompts vs retrain models vs investigate deeper.

Hou C.
Hou C.
Mental Models for Production AIMental Models for Production AI
Systems That Don't Learn, Decay

Systems That Don't Learn, Decay

Closing the loop. Why continuous learning isn't optional—systems that don't improve, degrade.

Hou C.
Hou C.
Mental Models for Production AIMental Models for Production AI
Silent Failures Make Me Most Nervous

Silent Failures Make Me Most Nervous

The degradation timeline. How AI systems fail slowly without anyone noticing—and how to catch it.

Hou C.
Hou C.
Mental Models for Production AIMental Models for Production AI
When Do I Need Human Review?

When Do I Need Human Review?

Where humans belong in the loop. An interactive framework for deciding when and how to involve human reviewers.

Hou C.
Hou C.
Mental Models for Production AIMental Models for Production AI
How Do I Know It's Failing?

How Do I Know It's Failing?

Detection before it's too late. An interactive guide to setting up failure detection for AI systems.

Hou C.
Hou C.
Mental Models for Production AIMental Models for Production AI
The AI Failure Cascade

The AI Failure Cascade

Every layer is a potential failure point. Maps the path from user request to user impact and where things can break.

Hou C.
Hou C.
Mental Models for Production AIMental Models for Production AI
Build Discipline vs Run Discipline

Build Discipline vs Run Discipline

Two different mindsets required for AI features. What it takes to build well vs what it takes to run well—they're not the same.

Hou C.
Hou C.
Mental Models for Production AIMental Models for Production AI
Build vs Buy: Vendor Strategy for AI Features

Build vs Buy: Vendor Strategy for AI Features

When to use APIs vs build your own. An interactive guide to the build-vs-buy decision for AI capabilities.

Hou C.
Hou C.
FirstBack
1234
Next Last