Tyler Wells

What I Learned Building a Multi-Agent Document Analysis System

Fri, 24 Apr 2026 00:00:00 +0000

[Multi-Agent Series 4/4] What worked, what broke, how chunking caused false conclusions, and what I'd do differently in v2.

Coordinating Multiple LLM Agents: Cross-Domain Synthesis

Thu, 23 Apr 2026 00:00:00 +0000

[Multi-Agent Series 3/4] How a coordinator turns specialist findings into cross-domain insights and decision-ready recommendations.

Building Specialist LLM Agents: Technical, Risk, Cost, and Timeline Analysis

Wed, 22 Apr 2026 00:00:00 +0000

[Multi-Agent Series 2/4] How specialist prompts, shared inheritance, and structured JSON outputs turn one LLM into multiple domain-specific analyzers.

Why Multi-Agent Systems Beat Single Agents for Complex Documents

Tue, 21 Apr 2026 00:00:00 +0000

[Multi-Agent Series 1/4] Why one large prompt breaks down on RFPs and contracts, and how specialist agents plus coordinator synthesis produce better analysis.

Building Ozark Ridge: Lessons Learned and What I'd Do Differently

Thu, 16 Apr 2026 00:00:00 +0000

[Retail AI Series 5/5] What worked, what didn't, what I'd do differently in v2, and why this project matters for e-commerce AI.

Building the AI Product Assistant: Context Injection, Multi-Turn Chat, and Cross-Product Retrieval

Wed, 15 Apr 2026 00:00:00 +0000

[Retail AI Series 4/5] How to build a Rufus-style AI assistant that answers product questions, suggests complementary gear, and builds camping loadouts — with conversation history, context injection, and dynamic retrieval.

Keyword Search vs Semantic Search: Why Natural Language Queries Need Vector Embeddings

Tue, 14 Apr 2026 00:00:00 +0000

[Retail AI Series 3/5] Side-by-side comparison of keyword and semantic search, why keyword search fails on natural language queries, and what the retrieval scores actually tell you.

Building the Catalog and Ingestion Pipeline: Archetypes, Embeddings, and ChromaDB

Mon, 13 Apr 2026 00:00:00 +0000

[Retail AI Series 2/5] How to generate 1180 realistic products from 20 archetypes, why product descriptions matter for RAG, and how the ingestion pipeline works.

Building AI Search for a Retail Website: The Stack and Why

Sun, 12 Apr 2026 00:00:00 +0000

[Retail AI Series 1/5] Building a mock outdoor retail site with AI-powered product search and a Rufus-style assistant. This post covers the architecture, stack decisions, and why RAG matters for e-commerce.

AI-Powered QA Testing with playwright-cli and GitHub Copilot

Thu, 09 Apr 2026 00:00:00 +0000

How to use playwright-cli with GitHub Copilot as an autonomous QA agent — without Playwright MCP, without writing test code, and without needing Copilot Vision or an embedded browser.

What I Learned Building a LangGraph Agent From Scratch

Mon, 30 Mar 2026 00:00:00 +0000

Building a LangGraph job research agent showed me where agent loops actually help, why typed state matters, and how conditional edges change the design.

Your MCP Server Is Only as Good as Its Docstrings

Sun, 15 Mar 2026 00:00:00 +0000

Building a college football data MCP server taught me that the most important design decision in an agentic system isn't the architecture — it's the docstrings.

How a Simple Power Automate Workflow Automated 250+ Hours of Work Per Month

Fri, 20 Feb 2026 00:00:00 +0000

Eliminate costs tied to repetitive corporate tasks with simple workflows like this -- a Microsoft Form + a Power Automate flow + an AI prompt

Scoring RAG Answer Quality with an LLM Judge

Mon, 26 Jan 2026 00:00:00 +0000

[RAG Series 3/3] Source URL retrieval tells you whether the right content was retrieved. It doesn't tell you whether the answer was any good. Adding an LLM judge to the eval harness reveals two failure modes that retrieval scoring alone can't see.

How to Design RAG Eval Test Cases

Sat, 24 Jan 2026 00:00:00 +0000

[RAG Series 2/3] How to write test cases that catch real retrieval problems, why source URL retrieval is a useful proxy metric, and when it isn't enough.

RAG Retrieval: Chunking, Embeddings, Reranking, and an Eval

Thu, 22 Jan 2026 00:00:00 +0000

[RAG Series 1/3] Covers chunking strategy, embedding model consistency, reranking, and building an eval harness — including what happened when Voyage AI's free-tier rate limits forced a more resilient architecture.

About

Mon, 01 Jan 0001 00:00:00 +0000

I’m Tyler, a software engineer with over a decade of experience building production systems. My focus has shifted — I spend most of my time now on AI automation and transformation: figuring out where AI can eliminate the repetitive, time-consuming work that slows organizations down, and then actually building the tools to do it.

In practice that means building AI agents, automating workflows that used to require manual effort, and creating reusable internal tools that give people back hours they were spending on work that shouldn’t require a human. The goal is always organizational velocity — fewer bottlenecks, faster execution, less time spent on the mechanical and more on the meaningful.