Developer-Grade Prompt Engineering
Standard conversational AI prompts fail at scale. I conduct extensive forensic research on LLM behavior to build indestructible XML State Machines. My custom AI engines eradicate hallucinations, enforce mathematical QA checks, and scale production output without sacrificing quality.
🔬 Forensic LLM Diagnostics
Crushing "Workflow Collapse"
The Discovery: LLMs suffer from "Eager Execution" when they hit roadblocks (like blocked web crawlers), panicking and hallucinating generic data to reach the final output format.
The Fix: Engineered strict "Anti-Panic Guardrails", URL Slug Hallucination blocks, and "Brake Pedals" that mathematically force the AI to halt and request human intervention rather than hallucinating.
ICAM & Invisible Math
The Discovery: LLMs suffer from "attention loss" in massive prompts and cannot natively run QA checklists or character-count constraints while generating text.
The Fix: I built ICAM to destroy the AI's "yes-man" bias, and use hidden <thinking> tags to force the AI to process constraint math and 15-point compliance rubrics invisibly before presenting the final output.
Defeating the Fluency Heuristic
The Discovery: AI detectors flag content because of perfect, sterile symmetry (The Grammar Halo Effect). LLMs naturally write in a mathematically predictable rhythm.
The Fix: I sabotage predictable grammar by enforcing Asymmetrical Jaggedness, banning LLM transition crutches, and mandating tactile, sensory practitioner evidence in the drafting pipeline.
🛠️ Custom XML LLM Engines
GBP Service Master Engine
Dual-mode routing (Full Build vs. Evaluator) with strict deduplication locks to automate Google Business Profile mapping. Generates 1,000-character, scannable descriptions mapped directly to core services.
The Meta-Prompt Builder
A highly advanced meta-prompt designed to write, evaluate, and harden other system prompts. Features a 14-point "A-N Evaluation Framework," automated Git-style DIFF reporting, and mathematically enforces an 80/100 Stability Score.
GBP Product Builder Engine
Transforms raw client domains into high-margin product listings using strict custom category rules and pricing structures. Eliminates the "No Page = No Product" error loop.
<operating_principles>
<forensic_anti_collapse_guardrails>
<!-- 1. WEB CRAWLER PANIC PROTOCOL -->
<rule>If you are blocked (e.g., 403 Forbidden),
you MUST explicitly admit the failure.
You are STRICTLY FORBIDDEN from guessing
or hallucinating generic industry services.</rule>
<!-- 2. THE BRAKE PEDAL -->
<rule>You are strictly forbidden from generating
the Final Output until all validations
and QA passes are explicitly completed.</rule>
</forensic_anti_collapse_guardrails>
<icam_mode>
<rule>INDEPENDENT CRITICAL ANALYSIS MODE:
1. Evaluate inputs independently.
2. Challenge risky assumptions.
3. Prioritize correctness and GBP safety
over human agreement.</rule>
</icam_mode>
</operating_principles>Programmatic Publishing Workflows
I build automated operations that connect API data extraction to programmatic LLM outputs. Using tools like n8n, Zapier, QuickClaw, and Apify, I design rigid SOPs that scale production bandwidth without sacrificing quality.
Automated Market Extraction
Engineered a Zapier/n8n pipeline to bulk-scrape competitor reviews via Apify, feeding raw JSON data into Claude to extract "Dealbreakers" and "Wow Factors" to inform the Master Client Dossier.
⚡ ICP Generation Logic ▼
The "You vs. Perfect" Blueprint: Automating the creation of a "Master Client Dossier," locking in upstream SEO targets and exact brand voice constraints before any content is written.
Zero-Click Defense: Forcing Claude to identify psychological hooks from raw data to make users bypass AI-generated search summaries.
Programmatic UX Web Copy
I build XML-based content pipelines that map copy perfectly to UI wireframes, trimming cognitive overload sections based on "Urgent Repair" vs "High-Ticket" intents.
⚡ Decision Clarity Logic ▼
Specifics > Claims: Explicitly forbidding LLMs from using subjective claims ("We are reliable"), forcing the generation of objective mechanisms ("Upfront Flat-Rate Pricing").
Section Goal Anchoring: Forcing the AI to output the psychological [GOAL] of a wireframe section above the copy to prevent semantic drift.
Autonomous SOP Pipelines
Converted human-led SOPs into executable AI state-machines. Automates high-converting UX wireframes for Regional Hub pages, and enforces "Jet Digital" scannability rules.
⚡ UX & GEO Alignment ▼
Jet Digital Standards: Banning 4+ sentence paragraphs and forcing "Key Takeaway" modules for mobile-first scannability.
AIO Snatch Blocks: Forcing "definition-first" generative engine optimization (GEO) beneath H2s for Google AI Overviews to extract.
Zero-Touch CMS Injection
I engineer direct API integrations that bypass manual data entry entirely. Utilizing custom webhooks, the WordPress REST API, and Headless CMS concepts, I force validated AI outputs directly into production databases.
⚡ Platform Architecture ▼
Backend Mastery: Advanced manipulation of WordPress, Webflow, and custom databases via API payload structuring.
Formatting Constraints: Forcing LLMs to output perfect HTML and XML tags to ensure CMS styling is perfectly preserved upon injection.
Technical SEO Architecture
Moving beyond basic on-page plugins, I engineer deep structural SEO protocols to feed conflict-free data to Google's Knowledge Graph, trigger Local Vision AI, and automate tedious technical QA processes.
Vision AI & The Landmark Play
Faking GPS EXIF data is obsolete. Google's algorithms now use Optical Character Recognition (OCR) and Vision AI to verify local relevance inside images. I engineered a specialized AI pipeline to generate service images featuring highly recognizable local city landmarks and localized OCR tags (e.g., municipal permit stickers) to mathematically prove geographical entity relevance on Regional Location Pages.
Master @graph Schema Protocol
Engineered a strict implementation protocol for advanced @graph JSON-LD schemas. Resolves complex multi-location entity logic to spoon-feed LLMs and RAG engines.
Architectural Mechanics:
- The Global Power Block: Injects the primary business entity into the Global Header to act as a unified, conflict-free Source of Truth site-wide.
- Lean Page Routing: Page-level schemas strictly reference the global entity using
"@id": "[[HOMEPAGE URL]]/#organization"to prevent duplicate entity generation. - Semantic Entity Mapping: Natively structuring LSI & NLP entities into the schema so AI instantly recognizes topical authority.
Algorithmic Content Audits
I designed the "3-Lane Audit Protocol", an automated triage system to diagnose SEO failures, protect existing link equity, and orchestrate the rescue of legacy content.
The Triage Process:
- Link Equity Triage: A mandatory ScreamingFrog/Ahrefs check ensuring that terrible content with strong backlinks is flagged for the "SEO Rescue" rewrite rather than deletion.
- FAQ SEO Rescue: Saving keyword density on legacy pages by migrating massive text walls into hidden, Schema-enabled accordion toggles for mobile UI conversion optimization.
📊 Technical QA & Logistics Automation
I bridge the gap between Technical SEO architecture and daily VA execution. Using Zapier and n8n, I route ScreamingFrog crawl data, Yoast configurations, and Google Search Console API alerts directly into Master Tracking Dashboards.
Performance Creative Strategy
Bridging the gap between raw performance data and high-impact creative execution. Managing high-volume Meta Ads ecosystems through full-funnel strategy, rapid A/B hook testing, and deep-dive consumer psychology.
Campaign Performance Snapshot
Deep-Dive Ad Diagnostics
Before testing new creatives, I diagnose the existing funnel. I map CTR, Hook Rate, Hold Rate, and CPA against industry benchmarks to uncover exactly why an ad is not scaling.
- Creative Fatigue Analysis: Structuring fresh pipelines when winning hooks degrade.
- Funnel Friction: Isolating landing page drop-offs vs. ad drop-offs.
- Ad Cap vs Ad Fatigue: Creating new visual angles to unlock new audience pockets rather than fighting the Meta algorithm.
Direct-Response Storytelling
Translating complex clinical data into highly structured, compliant visual references and long-form copy.
- Navigate strict Meta advertising constraints by developing compliant marketing mechanisms like "Bacterial Burnout".
- Utilize cognitive bias triggers (Specificity Bias, Negation Bias, Humor Effect) to bypass the buyer's guard.
- Develop detailed execution briefs mapping core emotions and medical literacy to awareness stages.
The POV Read-and-Loop
Leveraging Meta's algorithmic preference for high watch time through text-heavy narrative.
Why it prints:
- Silent Consumption: Requires zero audio to understand, capturing the 70%+ of users scrolling on mute.
- Native Aesthetic: Looks exactly like organic platform content, lowering the user's "ad blindness" defenses.
- High Replay Rate: The looping background video forces the user to watch multiple times just to finish reading, triggering the algorithm to prioritize the ad.
The 4 Winning Static Frameworks
