Economics student building revenue systems. GTM Engineering · Growth · RevOps, Athens & Santorini, Greece

[email protected] · linkedin.com/in/alexandros-nomikos-297ba538b

I'm a second-year Economics student who builds the systems, not just the spreadsheets.

I use AI tools to ship real products, but the architecture, pricing, and trade-offs are mine. What I bring is the combination: I think in unit economics and capacity utilisation, and I can build the system that produces those numbers instead of waiting for someone else to build it.

Education

BSc Economics — Athens University of Economics and Business (AUEB). Second year, 2025–

Selected work

Outreach Sequencer — GTM Infrastructure Tool

Builder · GitHub · outreach-sequencer-pi.vercel.app Next.js · TypeScript · Python · Prisma · PostgreSQL · Mailgun · Clay · HubSpot · Claude Code

Built my own outbound sequencer to run my job search as a GTM campaign. Multi-step timed sequences, per-timezone send windows so each email arrives in the recipient's morning rather than mine, bounce handling, daily and per-run send caps, and full HubSpot CRM sync (contacts, deals, email logging, pipeline stage auto-advancement). Contacts enriched through Clay. Paired with a pipeline that scrapes ATS job boards across three providers, classifies the companies and ranks them on why they are worth contacting now — 7,190 postings processed. The first batch showed zero opens and I assumed a deliverability failure. The Mailgun logs disproved it: every message was accepted, DKIM and TLS passing, one recovered from the recipient's inbox. Plain text cannot carry a tracking pixel, so open rate was never measurable — the metric was broken, not the delivery. I replaced it with bounce rate, which is the number that can actually damage a sending domain, and it stands at 0%. Being wrong in public about the cause taught me more than the build did.

I later added a verification pass that re-fetches every cited job posting live and refuses to send a claim it cannot source. It caught four false statements, and one case where a guessed ATS slug had attached a different company's job board to a contact, before any of it shipped.

Nommar — custom booking engine & RevOps foundation

Founding builder · nommar.gr — live August 2026 · Next.js · Prisma · PostgreSQL

A wellness spa opening in Santorini with two therapists, two rooms, a hard deadline and nothing else. I built the booking engine from scratch, parallel multi-guest reservations validated against real therapist and room availability, bilingual, with a full email lifecycle and self-service cancellation that returns slots to inventory. On the data side I caught a flaw that would have silently rewritten the value of every past booking whenever a promo changed, and rebuilt pricing as an immutable per-appointment snapshot; adding no-show/completion tracking then surfaced two latent bugs, including a double-booking risk. I also set the pricing: island-wide market research found the signature treatment had no competitor on Santorini, and a cost model put break-even at ~2 treatments/day with each additional daily treatment worth ~€16k per season.

Results — bookings, AOV, no-show rate, utilisation — from September 2026.

AI Hotel Concierge

Proof of concept

A backend that reads incoming hotel guest email and prepares a contextual reply as a draft for human approval, the person stays in the loop by design, because an automated wrong answer to a paying guest costs more than a five-second review. Modelled to cut drafting time substantially; validating it against real inbox volume is the next step. A related version runs in production for Nommar.

Freelance