Developing a line of internal corporate portals:

• Infrastructure & Operations Portal — management of company infrastructure (VMs, servers, VPN, access, secrets) and communications/HR (documentation, news, feedback, reporting). About 20 modules in one product, constantly growing.

• Service Management Portal — external product for a large corporate client in the ITSM domain: formalized interaction between operations (Ops/SRE), development, support, and management. Role-based interfaces on top of standard ITSM entities (services, incidents, changes, releases, transition, risks).

We need a lead developer who will take ownership of the products: design the architecture of new modules, write code, be responsible for stability, security, and performance. The team is small, with little bureaucracy — this is a role with real responsibility and broad autonomy, not a "ticket executor".

Process feature: a significant portion of the code is written through Claude (Claude Code / API). The candidate is a person who leads development with an AI agent as the primary executor: formulates tasks, reviews results, maintains the architecture and product meaning.

---

Responsibilities

• Design and implement new modules (backend + frontend + DB migrations + cron jobs + integrations).
• Support and develop existing ones: refactoring, eliminating technical debt, optimizing queries and frontend.
• Integrate with external systems via SSH/HTTP/API (control panels, billing, messengers, helpdesk, hypervisors; for the ITSM portal — monitoring, CI/CD, ticketing/CRM).
• Work with security: SSH keys, secret encryption, RBAC, audit logs.
• Support development infrastructure: Docker environment, migrations, deployment, monitoring.
• Lead development through Claude Code: decompose tasks into requirements understandable by both humans and LLMs; conduct code reviews of everything generated by AI; catch "works, but crudely"; roll back unsuccessful iterations.
• Conduct reviews, mentor, document architectural decisions.

---

Technology Stack

Backend — PHP 8.x (without heavy frameworks — pure PHP + custom router and DB layer) Database — MySQL / MariaDB, PHP migrations Frontend — Alpine.js 3, Tailwind CSS (without build tools and Node pipeline) Realtime — Server-Sent Events, BroadcastChannel Infrastructure — Docker / docker-compose, Linux (Debian/Ubuntu), cron, systemd, SSH Cryptography — OpenSSL (AES-256-GCM), libsodium (Ed25519), DEK/KEK External Systems — libvirt, OpenStack, OpenVPN, Rocket.Chat, Redmine, monitoring, CI/CD tools, ticketing AI Tools — Claude Code / Claude API as the primary code executor

The main thing is the ability to work without the "magic" of frameworks. The project does not use Laravel/Symfony/Vue/React — everything is done with explicit calls, closer to the roots of PHP and JS. This, by the way, makes the codebase convenient for AI: fewer hidden conventions, more explicit code that Claude reads and writes well.

---

Requirements

Mandatory

• 5+ years of commercial development in PHP, project leadership experience (not just writing code, but also making architectural decisions).
• Practical experience working with AI agents (Claude Code, Cursor, Copilot Workspace, Aider, similar) — not just "tried it", but led real projects through them for at least 6 months. Understanding how to write requirements for LLMs, how to review AI-generated code, where AI makes mistakes most often.
• Proficient SQL (MySQL/MariaDB): schema, indexes, query optimization, migrations, transactions, locks.
• Experience working without frameworks or with minimal ones (Slim, Yii-style, custom routers) — should be able to easily keep the entire stack in mind, from routing to templates.
• Frontend: proficient vanilla JS + experience with Alpine.js, Vue.js, or Knockout.js (any reactive JS without a build tool).
• Tailwind CSS or another utility-first CSS framework.
• Linux at a strong user level: bash, SSH, cron, systemd, permissions, sudoers, log reading.
• Docker and docker-compose: building images, debugging containers, multi-service compose.