An LLM-powered web app that rewrites LinkedIn profile content for tech workers between jobs. Free, no signup. 100+ sessions in the first 48 hours of launch from organic distribution.
What I shipped
- A prompt-driven rewrite engine across four content modes (headline, bullet, summary, post) with three-level editing intensity and a diff-based review UX so users can see exactly what changed and why.
- Direct REST integration against Anthropic and Gemini APIs (no SDK) with primary/fallback model chains, jittered exponential backoff, and a paid-tier provider posture so user content stays out of training corpora.
- A custom prompt evaluation framework with hand-curated cases hitting the live API. Eval-driven before there was a single user.
Operating the LLC
Facet ships under EiYa LLC, my own legal entity. That includes formation, terms of service, privacy policy, provider compliance review, and the production stack: Next.js, Vercel, Upstash Redis, Sentry with PII scrubbing, PostHog.
Post-launch positioning pivot
I launched it as a generic LinkedIn optimizer. Within the first week, the usage signal told a different story: the people coming back were tech workers between jobs, looking for narrative help during transition. I removed the signup gate, repositioned the landing, and the site now reflects the audience it actually has.
What this demonstrates
Operator-builder identity in evidence: end-to-end LLM product shipped under my own LLC. LLM provider strategy and posture in production, not as a slide. Eval frameworks I built personally, not specced for someone else to build. Product positioning agility from real usage data, not the launch deck.