I ship LLM products from idea to production.
Applied AI & full-stack. I take LLM products from idea to production — and build the reliability layer that makes them hold up.
four 0→1 builds · Peraton Labs research · engineering since 1996Founder or lead engineer on each.
The anti-drift framework behind how I ship AI-native software: an MCP server that gates autonomous coding behind hard, blocking quality checks — with a 7-phase workflow ledger and persistent memory on a Rust engine substrate. The gated Ralph loop that keeps agents on-task instead of drifting; the reliability method the other builds run on.
2–4×
AI-search visibility engine: continuous monitoring pipelines — controlled LLM querying, crawlers, analytics — that detect model-response drift, competitive displacement, and visibility loss. Drove 2–4× AI mention coverage in controlled deployments.
1,600+
Applied-AI SaaS that generates, validates, and scores business ideas on market, compliance, and feasibility — OpenAI Responses API, multi-stage pipelines, pgvector semantic retrieval. 1,600+ users, validated by paid conversions.
Pre-transaction fraud detection for SaaS: client-side signals, real-time analytics, and server-side logic that catch velocity spikes and abnormal interaction patterns before downstream processing.
Public writing on building LLM products — applied AI, retrieval, evaluation, and the systems that keep them reliable in production.
ChatGPT, Perplexity, Claude, Gemini, and Google AI Overviews each parse JSON-LD differently. The deltas matter. Here's what I've measured running the same @graph through all five.
Field-level JSON-LD bugs that fail silently across 45K+ PDP catalogs. Each one mapped to the rich result it kills, the AI engine that drops you, and the dollar cost I model for clients.
Production schema goes invisible to AI engines on its own, with no obvious failure. Here's the regression timeline I've measured across client engagements, and why CI gates aren't optional.
New writing on shipping LLM products — applied AI, retrieval, evals, and the reliability layer. No cadence promises, no spam. Unsubscribe anytime.
Three ways to bring me in, matched to where you are.
Embedded on the team, shipping LLM products from idea to production — with the full-stack rigor to keep them working once real users arrive.
A scoped build, idea to production, by an engineer who has taken four products 0→1. One clear deliverable, one clear timeline.
Reliability layer, anti-drift, evals, and workflow design — the parts that decide whether an agentic system is trustworthy enough to ship.
If you're shipping an LLM product — or you need the reliability layer that keeps one working in production — that's the work I do. The fastest way to reach me is right here.
engineer to engineer — no recruiters, no decks