dok.sh — Portfolio and CV Platform

Summary

dok.sh is the site you are reading. It replaces a static CV with structured content in English and Bulgarian, edited in an admin area. The same data produces the public pages, PDF CVs in three variants and two designs, the pages search engines index and the plain-text files AI assistants read.

Subtitle

One set of data for a bilingual website, PDF CVs, search engines and AI assistants.

Domain

Portfolio / Content platform

Target users

Potential clients, recruiters and hiring managers looking at my experience and availability.

Role

Founder / Product Engineer

Challenge

My CV, LinkedIn profile and old website said slightly different things, and none of them showed how I work on a project. I wanted one place to keep everything current in two languages, that also produces a CV people and screening software can read, and pages I can link to from LinkedIn.

What I owned

Content model, backend and frontend, admin editors, translations, PDF CV generation, SEO and AI discovery, deployment on a shared server, backups.

Approach

Structured data first, pages second; English and Bulgarian from the first commit; search engines and AI assistants treated as visitors; the public side kept simple and the complexity kept in the admin.

Outcomes

One edit updates the site in both languages, the PDF CVs, the sitemap and llms.txt. Content is prepared locally, checked as a YAML file and loaded on the live site in one step. Twenty use case pages about my two products were published this way in two days.

Details

Overview

I built dok.sh because my CV, my LinkedIn profile and my old website kept drifting apart, and none of them showed how I work on a project. The site is also a working sample: it is built the way I would build a client's admin-driven application, and it is updated the way I would run one.


Structure first

Everything on the site is structured data: positions, projects, case studies, use cases, services, skills and settings, each with English and Bulgarian text. Pages only display that data. A project can appear as a card, a detail page, a case study and a list of use cases without being written four times.

flowchart TD
    V[Visitor] --> W[Angular site, EN and /bg/]
    B[Search engine, AI crawler, link preview] --> N[nginx]
    N --> W
    N --> R[Server-rendered page]
    A[Admin] --> AD[Angular admin]
    W --> API[Spring Boot API]
    R --> API
    AD --> API
    API --> DB[(MongoDB + GridFS)]
    API --> PDF[PDF CV generator]
    API --> TXT[Sitemap, llms.txt, Markdown]

One source, several outputs


PDF CV with a page budget

Each CV variant has a target number of pages. The PDF is rendered and measured; if it is too long, it is rendered again one step denser, with older roles in less detail, until it fits. The ATS design uses the same content in one column with no tables, and both designs carry a title, author and skill keywords in the PDF metadata.

sequenceDiagram
    participant V as Visitor
    participant API as Backend
    participant CV as CV builder
    participant PDF as PDF renderer
    V->>API: Download CV (variant, language, design)
    API->>CV: Build from resume, skills and CV settings
    loop until it fits the page budget
        CV->>PDF: HTML
        PDF-->>CV: PDF and page count
    end
    API-->>V: PDF file

The admin

The admin was redesigned in November 2026 on Angular 22 and PrimeNG 22. Every collection uses the same layout: a list with filter and drag-to-reorder beside the selected item's form. One save bar per page shows unsaved changes and saves with Ctrl+S; leaving, reloading or logging out with unsaved changes asks first. Ctrl+K searches all admin pages. Images are dropped or pasted where they are needed and kept in one media library; pages ask for the width they need and the server resizes from a fixed list of sizes.


Hosting

The site runs in Docker Compose (MongoDB, the Spring Boot API and nginx with the Angular build) on a small VPS. A separate Caddy container in front handles HTTPS for this site and two other projects on the same server, so each project can be deployed without touching the others. Full database backups are taken from the admin before each deployment.


Decisions that mattered

Two languages from the start. Every text field has an English and a Bulgarian value, and the language is part of the address. Adding the second language later would have meant changing every model and screen.

Search engines and AI assistants are visitors too. Most clients search for the service, not the name, and more of them now ask an AI assistant. Both get the same facts as people, from the same data.

Content that can be moved and reviewed. The site text is one YAML file. New content is prepared and checked locally, then loaded on the live site in one step, and a new server starts from the same file.

Public pages simple, admin pages capable. Visitors see a clean site. Ordering, visibility, layout, CV settings and editors live in the admin.


Use cases

Ten pages describe single features of this site in detail: all use cases of dok.sh.


Next steps

Technologies

Java 21, Spring Boot 4, Angular 22, PrimeNG 22, TypeScript, MongoDB, OpenHTMLtoPDF, nginx, Docker, Caddy

Engagement

PERSONAL_PRODUCT

Timeline

2026-03 – Present

Use cases