Raffy Nikoghosyan

Live, in daily use

Case study · Real estate

BulGhar Homes

Esc

What it is

BulGhar Homes is a real estate agency that sells homes on the Bulgarian Black Sea coast, in Burgas and Varna, and presents partner projects in Dubai. I lead marketing and sales and own the technical stack, in a team of three: the website, the CRM with its iOS app, the cloud server and the publishing engine.

Role
Managing Director and Sales Manager; owner of marketing and the technical stack.
What I did
Ran sales and marketing, and designed and built the website, the CRM with its iOS app, the cloud server and the publishing engine.
Result
Live in 10 languages. As of 26 October 2026: 5,169 posts published to Facebook groups across a register of 549 groups. As of October 2026 the Eurovision 2027 campaign had contacted 83 hotels and properties and recorded 14 replies.
Status
Live, in daily use
Period
2025 to today

What it does

What the platform does

  • 01

    Multilingual website

    A catalogue in several languages with smart search: filters by city, type and price, and a map.

  • 02

    New builds catalogue

    Projects under construction in Burgas, grouped and searchable.

  • 03

    Enquiries and forms

    Enquiry, viewing and sign-up forms with bot protection, straight into the CRM.

  • 04

    Private shortlists

    A client receives a selection of properties made for them through a private link.

  • 05

    CRM

    People, deals, viewings, an activity log, reminders and notifications, with access by role. As of 26 October 2026 it manages 58 contacts and 48 deals.

  • 06

    Native iOS app

    A new listing with photos and GPS from the phone, voice notes turned into text, and approvals.

  • 07

    Scheduled publishing

    Posts go out to Facebook, Instagram and groups at the minute set in the calendar.

  • 08

    AI copy in several languages

    Listing copy is drafted by AI and reviewed before it goes out.

  • 09

    Newsletter and morning email

    A newsletter for subscribers and a morning email called Today: viewings, waiting clients, tasks.

  • 10

    Self-checks and backups

    Every night the server reconciles its data and keeps encrypted backups for 35 days.

In numbers

  • 172HTML pages, plus 670 blog files
  • 65CRM tables
  • 218Swift files, about 46,400 lines
  • 79server modules, about 27,600 lines
  • 24custom AI assistants
  • 1,778commits

Context

The agency started work in July 2025 with a small team. It sells homes in Burgas and Varna and works with international buyers, so every listing and article needs several languages and daily visibility on social media.

The challenge

Compete with established agencies without a marketing department or an IT team. Leads came from portals, chats and phone calls, content had to go out every day, and numbers in public posts had to be correct.

What I built, layer by layer

  1. Website and SEO: a multilingual site with its own design system, technical SEO and a translation workflow that checks every language before release.
  2. CRM and iOS app: people, deals, viewings, the activity log, the property register and the calendar in one database, with a native Swift app for the team.
  3. Cloud server on Cloudflare: scheduled jobs every minute and every five minutes, 76 database migrations, nightly self-checks, encrypted backups and data stored in the EU.
  4. Marketing engine: carousels, short video and stories, a register of Facebook groups and a log of every publication, all driven by one calendar.
  5. AI assistants: custom assistants for listing copy, research with source checks, contracts from approved templates and translations. They draft; the team reviews.
  6. Design system: tokens, components and carousel templates, so content stays consistent without designing from scratch.
  7. Verification: every number in a public post is checked against its primary source, every language passes an automated check, and every release goes through a check before publishing.

What I learned

  1. Put the data in one place first. Every later feature became easier once the server was the single source of truth.
  2. Automate the boring checks, not the judgement. Machines reconcile and draft; people decide what goes public.
  3. Verify numbers at the source. A wrong figure in a post costs more trust than a late post.
  4. Small teams need systems even more than large ones, because nobody has time to repeat work.

Stack

  • JavaScript
  • Cloudflare Workers
  • SQL
  • Swift
  • Python
  • HTML and CSS
  • AI agents

Interested in working together?

Email or LinkedIn is the fastest way to reach me. My CV is one click away.

References available on request.