Sangmin Lee
AI Product Engineer. Coding since 1996. Founder of RankLabs. Writes engineering notes on building LLM products — applied AI, retrieval, evaluation, and production reliability.
The arc
I started coding in 1996, age 12, picking up Visual Basicwhile sitting in on my father’s MIS coursework during his MBA. JavaScript followed around 1998. The engineering DNA isn’t a pivot. It’s the foundation.
In 2008 I co-founded an IT consulting company with a college buddy and ran it for nearly seven years. The company pulled me away from full-time production code into stakeholder work, MVP scoping, Agile/SCRUM delivery, and release management. I picked up a masters in computer science in the middle of it. By the time we wound the company down in 2015, I was hungry to be back in the codebase full-time.
That’s the year I joined Complex Networks, a top-tier US digital media publisher. Partnered with their BI team across the sister-site portfolio (millions of monthly visitors per site), led the frontend revamp on three sister sites single-handedly before they consolidated into one platform. Engineering at publisher scale: BI pipelines, performance, and the systems that decide what gets seen — a decade before AI search was a category.
After Complex I spent five years deep in full-stack at Verdocs (e-signature platform, library-first architecture, framework-agnostic embeds). Then four-plus years as a Senior Research Scientist at Perspecta Labs (which became Peraton Labs after the 2021 acquisition), building Cesium-based 3D geospatial visualization for federal contracts. Compliance-heavy environments. Real engineering rigor.
I founded Mindcraftor in 2025 (applied-AI SaaS for idea generation, scaled to early adoption then shut at zero MRR, a useful failure). I founded RankLabs later in 2025, and RecoverPay after that — each an LLM product taken from idea to production. I also build OpenSquid, the anti-drift reliability layer for agentic coding.
I currently consult through SAE.SO on two parallel tracks: ongoing A/B testing / experimentation work for a Fortune-1000 health & wellness brand (300+ tests and counting since 2021), and a senior full-stack engagement on a separate end-client (2026-02 onwards). The experimentation muscle stays current. I’m still shipping in production every week.
RankLabs is my primary independent practice. I take a small number of 0→1 engagements per quarter. I build and ship the product myself — idea to production — and I don’t take on work I can’t deliver.
Where the work happened
RankLabs
Founder
AI visibility infrastructure: proprietary crawler, structured-data extraction, page-level scoring, AI engine emulation, field-level schema validation.
2–4× AI mention coverage in controlled deployments. Used in every client engagement as the diagnostic backbone. Several capabilities don't exist in incumbent tools (Screaming Frog, Botify, Sitebulb, Onely).
Mindcraftor
Co-Founder · Senior Full-Stack
AI SaaS for idea generation, validation, and autonomous CX workflows. Built with Next.js, Node, Supabase, Stripe, RAG-style flows.
1,600+ users (paid-validated). Scaled to early adoption then shut at zero MRR. Useful failure.
SAE.SO
Senior Frontend Engineer (Consultant)
Long-running consultancy contract. 300+ client-side experiments delivered (and counting) for a Fortune-1000 health & wellness enterprise. A/B testing infrastructure ongoing. Concurrent senior full-stack engagement on a separate end-client (2026-02 onwards).
Five years and counting in production CRO/MVT. The experimentation muscle stays current.
Perspecta Labs / Peraton Labs
Senior Research Scientist
Started as Perspecta Labs; renamed Peraton Labs after the 2021 acquisition. Cesium-based 3D geospatial visualization for federal/compliance-heavy environments. Frontend modernization with lazy-loaded modules, modular libraries, state management.
Cut Cesium app load times by ~38% with Web Workers + 2D fallback. Mentored juniors. Set frontend standards across multi-million-dollar contracts.
Verdocs
Senior Full-Stack Engineer
E-signature platform. Pioneered library-first architecture: modularized e-signature flows into reusable libraries. Framework-agnostic embeds (Web Components / Angular custom elements). Mentored devs on scalable UI practices.
Complex Networks
Lead Front-End Developer
Top-tier US digital media publisher. Led the frontend revamp on three sister sites single-handedly before consolidation. Partnered with the BI team on cross-site performance and conversion experimentation. High-traffic publisher engineering at scale.
Frontend + data engineering at publisher scale — a decade before AI search was a category.
Computer Pathways
Co-Founder / IT Consultant
Co-founded with a college buddy. 6 years 9 months. Worked closely with business stakeholders to direct technology development across multiple enterprise teams: MVP scoping with backlog alignment, Agile/SCRUM delivery via JIRA, release management through demos, technical guides, and training. Translated complex technical content into stakeholder-readable form. Earned a masters in computer science in the middle of the run.
Articles by this author
- // 2026-05-02How AI answer engines actually resolve your @graph
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.
Read the article → - // 2026-05-0210 JSON-LD bugs I find in every commerce audit (and what they cost)
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.
Read the article → - // 2026-05-02Schema decays. Here's the half-life I've measured.
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.
Read the article →
Building something that has to hold up?
I take LLM products from idea to production, and build the reliability layer that keeps them working — the same anti-drift engineering behind OpenSquid. One engineer, idea to ship.
engineer to engineer — no recruiters, no decks