• Role

  • Founder & CEO of TRAIL Labs — designed, built, and launched TRAIL Search as the sole engineer, taking it from first prototype to production.
  • Owned the full stack: Python/FastAPI backend, TypeScript/Next.js frontend, evaluation, integrations, and operations across reliability, latency, and cost.

• What it does

  • Diagnoses how a brand is found and cited across AI answer engines (answer and generative engine optimization, AEO/GEO) and prescribes concrete fixes.
  • Runs recurring, per-engine tracking and produces prescriptive reports, so teams see not just a score but the specific bottleneck to fix next.

• Evaluation layer

  • Built evidence-graded scoring, factuality gates on LLM-generated drafts, and reproducibility checks over representative question sets.
  • The core engineering problem is turning noisy, shifting model outputs into decisions a team can act on, measured with representative data and production signals rather than one-off impressions.

• In-house measurement pixel

  • Developed a proprietary first-party pixel that customer sites embed to monitor AI-engine-driven visibility and traffic as a production signal.
  • Tracks signals per engine (never aggregated, so a drop in one engine is never masked by a rise in another), with daily drop detection and contact-event logging.

• Research

  • Studied the “content boundary” of generative search: how similar or different a page must be to what is already said in order to be cited, and how fact-level relations drive citation (realized marginal information contribution).
  • This work is under submission to WWW 2027 and ECIR 2027.