Alex
Story

Building practical AI systems,
web platforms & tools that
make complex work easier.

I'm a full-stack developer focused on applied AI, RAG systems, agentic workflows, CMS architecture, UI/UX thinking, and automation. This site is my personal knowledge hub — work, field notes, custom tools, MCPs, experiments, and lessons from building real systems.

Explore Work →Read Field Notes →
The full system — how everything connects
The proof
Work →
RAG knowledge platform
Agentic CMS pipeline
The thinking
Field Notes →
Tuning retrieval for messy PDFs
When agents should stop looping
The uniqueness
Toolbox →
Website-analyzer MCP
Eval-runner skill
The movement
Lab →
Model A vs B on extraction
Embedding drift experiment
Direction

What I'm Building Toward

My work is moving toward practical AI systems: tools that understand context, support real workflows, improve decision-making, and make complex technical processes easier to use.

I'm interested in the full system — not only the model, not only the interface, and not only the code. The value is in how everything connects: data, APIs, UX, automation, AI logic, evaluation, and the user journey.

Focus Areas

Core Focus

◈

Applied AI

RAG systems, agents, prompt workflows, document processing, AI evaluation, and structured LLM outputs.

⬡

Full-Stack Systems

Frontend, backend, APIs, CMS architecture, data workflows, and product platforms.

◐

UI/UX Thinking

Designing interfaces that guide users, reduce confusion, and make technical systems easier to understand.

⟲

Custom Workflows

Skills, MCPs, prompts, automations, and repeatable systems for building and analyzing software.

Selected Work

The Proof

Projects, systems, and prototypes that show how I turn complex technical ideas into usable workflows.

View all work →
12SYSTEMS SHIPPED
6DOMAINS
30+WORKFLOWS
4IN PRODUCTION
AllAI & RAGCMS & WebData QualityUI/UXAutomationDev Workflows
AI & RAGProduction
RAG knowledge platform
A context-aware retrieval system that answers from messy internal docs with traceable sources.
ProblemTeams couldn't trust answers pulled from scattered, low-quality PDFs.
BuiltChunking + hybrid retrieval + eval harness + cited-source UI.
ProofGrounded answers with confidence and source links, not guesses.
Next.jspgvectorLLM eval
Dev WorkflowsStable
Repo review automation
CI step that runs structured code + UX reviews and posts findings to PRs.
CIMCP
UI/UXPrototype
AI confidence UI kit
Patterns for showing certainty, sources, and fallbacks in AI-driven interfaces.
design system
Data QualityProduction
Document intelligence tool
Extracts structured fields from documents and flags low-confidence rows for review.
OCRschema
CMS & WebStable
Agentic CMS pipeline
Content moves from draft to publish through an automated review + approval flow.
Sanityqueues
Field Notes

The Thinking

Field notes from building, testing, comparing, breaking, and improving AI systems, web platforms, UI/UX flows, and developer workflows.

All field notes →
AllApplied AIRAG & AgentsUI/UXCMS ArchitectureLessons Learned
Featured · RAG & Agents8 min
What actually breaks retrieval on messy PDFs
A walk through the failure modes I kept hitting — and the chunking + eval changes that fixed them.
Takeaway: Retrieval quality is a data-cleaning problem before it's a model problem.
CMS Architecture
Modeling content for AI + humans
Schema choices that keep both editors and models happy.
◷ 7 min▲ Intermediate
UI/UX
UX patterns for AI confidence
How to show uncertainty without scaring users.
◷ 6 min▲ Beginner
RAG & Agents
When agents should stop looping
Cheap stop conditions that beat clever planning.
◷ 5 min▲ Intermediate
Toolbox

The Uniqueness

Custom skills, MCPs, workflows, prompt systems, and reusable tools I create to improve how I build, debug, review, and work with AI.

Open full toolbox →
AllMCPCustom SkillPrompt SystemWorkflowAI Evaluation
~/toolbox/featuredActive
CUSTOM SKILL · UX/UI REVIEW SYSTEMStable
UX/UI Review System
Feed it a page or component; it returns a structured design + usability critique against my rubric.
WORKFLOWBeta
Repo → release flow
commits → notes + deploy
PROMPT SYSTEMStable
Prompt system library
task → structured prompt
AI EVALUATIONActive
Eval-runner skill
cases → score report
MCPStable
Website-analyzer MCP
site → audit.json
Lab

The Movement

A lab of tests, comparisons, prototypes, and unfinished ideas that show what I'm learning next.

Open the lab →
Running 2
Prompt evaluationRunning
Structured-output prompt set
QDoes schema-first prompting cut retries?
Model comparisonRunning
Model A vs B on extraction
QWhich model extracts tables more reliably?
Completed 1
UI pattern testCompleted
Streaming UI prototype
ResultToken streaming felt faster even at equal latency.
Failed 1
AI workflow testFailed
Embedding drift experiment
ResultRe-embedding on every edit was too costly for the gain.
About Alex

The Personal Context

I'm a full-stack developer focused on practical AI, modern web platforms, CMS architecture, UI/UX thinking, and automation.

I care about how systems are structured, how users move through a product, how data flows, how AI decisions are guided, and how complex tools can become easier to use.

More About Me →
Let's Connect

Reach Out

For questions, collaboration, feedback, or technical discussion, you can contact me through the form.

Ask this site

Answers come only from the published work here, with sources cited.