Operating at the intersection of AI research and creative execution, this work spans foundational LLM investigations (transparency, alignment, RAG) and the practical business applications of multimodal generative AI. I design and build projects utilizing text, image, audio, and video generation, demonstrated through case studies in AI-driven branding, sophisticated video production, custom visual asset creation (leveraging tools like ComfyUI, Midjourney, and LoRA training), and complex knowledge structuring. This portfolio highlights a blend of technical depth, e-commerce functionality, and design expertise across the AI landscape.

<aside> <img src="/icons/book_orange.svg" alt="/icons/book_orange.svg" width="40px" />

Learn more by visiting https://whatsoever.ai

</aside>

LLM Research & Workflow Development

Key research projects and framework developments include:

TPIP (Truth & Pattern Inference Protocol) – A structured evaluation methodology for assessing LLM honesty, transparency, and epistemic integrity. TPIP enables comparative analysis of leading models by quantifying tension across helpfulness, harmlessness, and honesty axes.

GENESIS Framework – A modular, metacognitive system designed to guide LLM output toward more human-like, emergent expression. It has been successfully tested across creative tasks (e.g. narrative, poetry, informal writing), consistently producing results that pass advanced AI detection systems (Quillbot, Sapling.ai, etc.) as fully human.

Transparency Indexing and Model Audits – Ongoing benchmarking of models including GPT-4.5, GPT-4o, Claude 3.7 Sonnet, Gemini 2.5 Pro, and Deepseek R1. Each is evaluated across TPIP metrics, data visibility, inference traceability, and alignment disclosure.

RAG-Lite SEO & LLM Consensus Mapping – Applied a retrieval-augmented generation (RAG-lite) framework to reverse engineer Etsy’s ranking algorithm. This multi-model swarm (Claude, GPT, Gemini, Deepseek, etc.) was structured into a semantic knowledge map, enabling Claude to simulate retrieval augmented reasoning. The project achieved a #1 organic Etsy rank for the term “yoga art,” validating a scalable method for LLM based SEO and metadata optimization.

This research portfolio serves both as a documentation archive and a prototype environment for frameworks supporting improved alignment, authentic creativity, and verifiable truth modeling across generative AI systems.

TPIP: Exposing Alignment Tension in Modern LLMs

GENESIS: Achieving AI Content Authenticity

ETSY RANKING VIA RAG-LITE SEO

CUSTOM NODES FOR COMFYUI METADATA EXTRACTION

Creative AI Portfolio & Case Studies

Explore detailed case studies covering diverse creative AI applications, including: video, visuals, audio, branding, UI/UX, and custom model training.

<aside> <img src="/icons/book_orange.svg" alt="/icons/book_orange.svg" width="40px" />

View complete creative portfolio on Behance: **https://www.behance.net/whatsoever**

</aside>

<aside> <img src="attachment:d8c13d9c-1ab3-45ee-9fef-2c4a142bd962:linkedin2.png" alt="attachment:d8c13d9c-1ab3-45ee-9fef-2c4a142bd962:linkedin2.png" width="40px" />

View Linkedin profile: https://www.linkedin.com/in/codihall/

</aside>