# Hello (: I live at the intersection of [[Digital Public Relations (PR)]] and [[Neural Information Retrieval (IR)]]. I'm currently VP of Technology at [Razor Sharp PR](https://razorsharppr.com), a boutique firm in Austin, TX. Most importantly, I'm a husband and father of two. ## Work ### AnomalyLab (coming soon) Full-stack web application (Python/FastAPI, TypeScript/SvelteKit, PostgreSQL/SQLAlchemy) I began developing in April 2024 that compresses weeks of client and market research into minutes to supercharge every client meeting; Finds the expertise that makes your clients unique; Detects relevant story opportunities before they go mainstream. If you're in PR, think of it as AI-augmented [[Newsjacking]]. Crawls websites (Scrapy), generates embeddings (OpenAI), models topics (BERTopic), and detects statistical anomalies (z-score) in GDELT's 100M+ daily news articles that signal story opportunities. Uses LLMs for client commentary modeling and reporting (Anthropic). Type-safe architecture (Pydantic + TypeScript) with domain-ish-driven design, layered architecture, Alembic migrations, Docker Compose deployment, Uvicorn/Nginx serving, JWT auth, streaming responses, and Slack integration (for notifications). Now with distributed task processing (Celery/Redis/Flower). 🚀 ### [The No-Nonsense Guide to Getting Cited by AI](https://www.razorsharppr.com/s/getting-cited-by-ai-guide-cc.pdf) Culminating research into how AI assistants decide what to cite when answering questions. Started as an internal analysis of PR and marketing metrics (e.g., SOM, AEO, LLMO, GEO) in February 2025, and by October 2025 it had evolved into a formal investigation of [[Neural Information Retrieval (IR)]] applied to content strategy. We [announced](https://www.morningstar.com/news/accesswire/1098370msn/razor-sharp-pr-unveils-industrys-first-framework-for-winning-ai-citations) it's official release on November 7, 2025. - Most PR / marketing people don't have the time to understand retrieval mechanics - Most ML engineers don't view their research through the lens of content strategy - The combination of these skill-sets is what makes our approach rare ### [[Home Lab]] My home server runs local LLMs, self-hosted services (Jellyfin, Plex, these notes), and AnomalyLab.ai in Docker containers (Nginx). Started as an idea for a consumer NAS build in November 2024, but by December it had evolved into a SMB applications server with 72 TB usable storage (due to double parity). Custom $3,000 TrueNAS SCALE system with a RTX 3070 and enterprise-grade features (ECC memory, RAIDZ2, 6x enterprise HDDs). Dozens of hours researching optimal builds led to a relatively balanced system; Similar specs from TrueNAS retail for $4,850 (without a GPU). ### [[Neural Information Retrieval (IR)]] Research After ChatGPT 3.5 went mainstream in late-2022, I was inspired to learn the basics of [[Semantic Search]] to improve my PR workflow: I clustered client content to understand what they can speak authoritatively about and improve journalist list building (can't we have lists that update themselves?). Today, understanding how IR systems actually work—data ingestion pipelines, query understanding, embedding representations, retrieval and reranking methods—reveals that these AI assistants are sophisticated retrieval-augmented systems, not magic black boxes. They blend parametric knowledge (what's encoded in training data) with real-time external retrieval, using principles that have guided search engines for decades. In my digital public relations work, this research helps me build content strategies that survive platform shifts. This research comes to life in AnomalyLab. ## What's Published Here These are (mostly raw) working notes documenting what I'm learning and building. If you're working on similar problems, you might find something useful. [[Obsidian Publish Page Components]] ## Contact **Personal:** - [email protected] - https://x.com/ethanpy3 **PR:** - [email protected] - https://razorsharppr.com - https://www.linkedin.com/in/ethan-young-ba495a148 **Dev:** - [email protected] - https://github.com/ethanpy3