## Eugene Panov – Senior Full Stack Developer | Automation & Microservices Specialist | PHP
Location: Limassol, Cyprus (CY)
LinkedIn: https://www.linkedin.com/in/evgenii-panov
GitHub: https://github.com/evpanov
GitHub: https://github.com/panov-id
Telegram: @ppaannoovv
Blog: https://panov.id/blog/
Languages: Russian (Native), English (B1)
---
## Summary
Senior Full Stack Engineer with 10+ years of experience building scalable web applications and modernizing legacy systems, mainly in payments and fintech. Strong expertise in PHP/Laravel backends and React frontends, with a focus on clean architecture, reliability, and performance. Experience designing and implementing microservices, high‑load APIs, dashboards, and internal tools for fintech, e‑commerce, and analytics. Actively uses AI to speed up delivery, refactoring, and documentation while keeping control over architecture and code quality.
Looking for new opportunities as Full Stack Engineer, PHP Backend Developer, ReactJS Frontend Developer (remote or Cyprus-based).
---
## Core Skills
- Backend: PHP 7/8, Laravel, Lumen, PHPUnit, MySQL, PostgreSQL, Redis, RESTful APIs
- Frontend: JavaScript (ES6+), React, HTML5, CSS3
- Architecture: Microservices, system design, database design
- DevOps & Tools: Docker, Git, CI/CD, Kafka, Vault, KeyCloak
- Domains: Payments, fintech, CRM, dashboards, data processing
---
## Experience
### Agent Manager – NEIGHBRO.PLACE & SOSED.PLACE
Jul 2026 – Present • Limassol, Cyprus • Remote • Self-employed
---
### Engineer – NDA
Mar 2026 – Present • Full-time
---
### Senior Full Stack Developer – Bizcombo Technologies Ltd
Dec 2025 – Mar 2026 • Limassol, Cyprus • On-site
- Design and implement automation workflows connecting external/internal APIs.
- Build custom modules and architect scalable microservices for high‑performance backend systems.
- Focus on maintainable code, observability, and reliability of the automation platform.
---
### Full Stack Developer – F&F WORLDWIDE SERVICES LTD
Mar 2023 – Jul 2025 • Limassol, Cyprus • On-site
- Developed, refactored, and supported an in‑house DIY CRM & Client Area for a fintech company.
- Implemented and maintained payment gateway integrations with providers (B2BInPay, VirtualPay, Skrill, Neteller, etc.).
- Built and supported trader gateways and integrations for MT4/MT5.
- Performed data analysis, investigated and resolved production issues, and handled operational tasks.
- Stack: PHP 7/8, Laravel, Lumen, MySQL, Redis, Kafka, Vault, KeyCloak, PHPUnit, Docker.
---
### Full Stack Web Developer – Reyo Media Cyprus
Apr 2022 – Feb 2023 • Limassol, Cyprus • Remote
- Developed, refactored, and supported parts of a DIY analytics dashboard.
- Implemented procedures and services to retrieve large reports from affiliate platforms via APIs and web scraping.
- Monitored and resolved errors using Sentry, created and tracked issues in Asana.
- Set up Dockerized environments for local development.
- Stack: PHP 7.2+, Laravel, PostgreSQL, Redis, PHPUnit, Docker, Asana, OneSignal, AWS.
---
### Full Stack Developer – Unlimit
Dec 2014 – Feb 2022 • Limassol, Cyprus • On-site
- Developed, integrated, and supported a DIY e‑commerce platform (AddStore.com), including payment integrations (PayPal, OnlinePay), text search powered by Elastic, and YML import (Yandex Market).
- Built dashboards for MetaTrader and tools for merchants, including a Merchant Interview Questions application.
- Developed and supported RESTful APIs for partners and business products such as “Card Issuing”, and a relationship manager for SmartVista.
- Worked with high‑load financial systems and complex integrations in the payment industry.
- Stack: Yii2, PHP 7.4+, Laravel, ReactJS, MySQL, Oracle, Elastic, SOAP, JavaScript, jQuery, HTML, CSS, Twig.
---
### Web‑Developer – Dengi Online
Apr 2011 – Dec 2014 • Saint Petersburg, Russia • On-site
- Integrated payment systems (mobile commerce, card payments, wallet and terminal systems) into the aggregation platform “Dengi Online”.
- Developed and supported integration modules to ensure stable operation.
- Assisted clients integrating with the aggregation system and supported logistics services aggregators.
- Stack: Linux, Apache, MySQL, PHP, HTML, CSS, JavaScript, jQuery, XML, SOAP, PGP, OpenSSL.
---
### Earlier Roles
- Web‑Developer – Metrika (Wholesale Trade/Import-Export) (Aug 2010 – Apr 2011)
Developed and supported internal web interfaces (frontend and backend) for product catalogues, Axapta‑based information systems, and payment order preparation tools.
Stack: Linux, Apache, MySQL, MSSQL, PHP, HTML, CSS, JavaScript, jQuery, AJAX, ExtJS.
- Web Developer – KIT Finance Investment Bank (May 2007 – Apr 2010)
Developed and supported intranet portals and internet websites, including modules for document exchange and internal tools.
Stack: Linux, Apache, MySQL, PHP4/PHP5, HTML, CSS, JavaScript, Prototype JS, ExtJS, AJAX.
- ASP Web Programmer – Piter Plus OOO (Internet Services) (Feb 2006 – May 2007)
Developed and supported online shop modules and billboard systems.
Stack: IIS, ASP, MSSQL, HTML, CSS, JavaScript, Photoshop.
- Webmaster – Efo OOO (Electronics, Components, and Semiconductor Manufacturing) (Jan 2005 – Jan 2006)
Maintained content (news, articles), processed images, and supported company web presence.
Stack: HTML, CSS, Photoshop.
---
## Education
Peter the Great St. Petersburg Polytechnic University – Technics
2000 – 2006 • Russia
---
## Selected Projects
- SOSED.PLACE – https://sosed.place
A local neighbour network: say a word to people nearby, match, talk — and let go. No profiles, no feeds.
- NEIGHBRO – https://neighbro.place
Good neighbours. Real moments.
- Waste Collection in Cyprus – https://www.instagram.com/waste.collection.cyprus/
A 100% crowdfunding project: we gather donations and pay honest people who collect trash from the streets of Cyprus €2 per bag.
- SACRED GARDEN – https://www.instagram.com/sacred_garden_cyprus/
Rave venue near the village of YOO, Limassol.
- NOROOF – https://www.instagram.com/noroof_club_cyprus/
Rave venue in a quarry near the White Stones.
- Psytican – https://psytican.com
Web application in the mental health / psychology domain.
- Paparad – https://paparad.bandcamp.com
Personal creative project (music).
---
## AI Assistant Setup
How I work with an AI coding assistant. Nine approaches, each with what it is and why.
0. Commands, not guessed words.
What: a message without a command is talk — the assistant answers in words and every edit is refused that turn. Work opens only on /go, and the answer starts with "understood as …". /commit and /push allow exactly one commit or push within ten minutes, never by a sub-agent. /q, /iq, /redo, /stop, /fail, /win, /mem cover questions, restating, stopping and the ledgers.
Why: hooks that guessed intent from words kept misreading — a question taken for an order, a picked option for a question. A typed command is read exactly; the session's first reply lists them.
1. Rules are enforced by hooks, not by memory.
What: a short core of rules loaded into every session, and small scripts on the assistant's events that refuse a step breaking a rule.
Why: a rule the model only remembers fails exactly when the context is longest. A refusal at the moment of the slip does not depend on attention.
2. "Done" needs proof.
What: no "done" without a run after the last edit; numbers are recounted by a script; claims about a library are checked by an experiment in a container.
Why: the most expensive mistake is the confident one. A check costs seconds; a wrong "done" costs a lost evening.
3. Every slip is written down, in my words.
What: a ledger of failures, each with the owner's verbatim reaction, grouped into classes and shown before every reply; a separate ledger of what the owner explicitly approved — a line appears only on my /win, in my words, and each one names the proof (a test with a planted break, a live run, a number), the post https://panov.id/blog/pobeda-eto-dokazatelstvo.html walks through all five.
Why: the same mistake comes back in a new shape. A list of my own patterns, in front of me every turn, catches the shape instead of the instance.
4. A second opinion before it reaches me.
What: a read-only judge agent checks every draft against the failure classes; a large piece goes through a panel of independent reviewers with a mandatory security lens, and every finding is first attacked by another agent.
Why: one reviewer has one set of blind spots. Findings that survive an attempt to refute them are the ones worth my time.
5. Checked against the task, not only against the rules.
What: at the end of each turn a cheap model compares what was done with my last message; a short message must be restated as "understood as …"; a change touching several files is first retold by a second agent.
Why: work that drifts to a neighbouring problem passes every quality check — it is good work on the wrong thing.
6. A working memory the assistant queries itself.
What: a small local database the assistant queries instead of grepping files: my messages and corrections, decisions, the rules, skills, docs and review notes, per project. It answers "was this already decided?", "where did I slip on this file?", and a new session starts from "where did we stop?" — the last turns in my own words.
Why: asking me what I already answered, or repeating a fixed mistake, costs more than a lookup. Command output never goes in, so no secrets.
7. A map of the project instead of searching it.
What: a generator — shared, or the project's own where its stack needs one — draws what links to what — code, database, documents — checks every link against the disk while building, and writes a diff to a journal when something changes.
Why: searching the code anew every session is slow and misses links. A map built from the project itself and checked on every build can be read in a minute and trusted.
8. Cheap work goes to cheap models; the decision stays with me.
What: mechanical steps with a machine check go to small executor agents; noisy output is read through a 12-line digest; I am asked only about money, accounts, irreversible actions and product forks — never about technique.
Why: tokens and my attention are the two scarce resources. Everything else the assistant decides, writes down, and reports with numbers.
Everything runs locally, nothing is installed on the host outside containers, and the setup lives in ~/.claude.