Open to relocation (EU) · remote-friendly

Software developer, AI-accelerated

I ship products end-to-end, working in tandem with AI coding agents (Claude Code, Codex): web apps, SaaS, internal systems, a native Android app, even ESP32 firmware. Deepest track record in PHP/Laravel — but the stack follows the problem, not the other way around.

Message me on Telegram GitHub

4
products shipped in the last 12 months: 13+
years in digital: ads, SEO, analytics: ×10
lead growth after one site rebuild: 24/7
my systems run unattended in production

What I do

Product development

From idea to production: architecture, data model, UI, deployment. Small, frequent, verifiable releases.

Internal systems & CRMs

Admin panels, process automation, integrations: search engines, messengers, imports, document generation.

SaaS & API integrations

OAuth flows, third-party API sync, background queues, multi-tenant data models.

Marketing-aware engineering

13+ years in ads and SEO mean the things I build are measurable by design: analytics, conversions, search-friendly structure.

Beyond the browser

Native Android app, ESP32 firmware, Telegram bots — I pick up new platforms quickly with AI tooling.

Tools I ship with

PHP 8
Laravel
Filament
Livewire
MySQL / SQLite
Redis
Meilisearch
Laravel Horizon
Docker
Tailwind CSS
Astro
Alpine.js
Android
ESP32 / C++
Telegram Bot API
OAuth 2.0
CI: tests, PHPStan, Pint
Claude Code
Codex

Selected work

hours → minutesof reporting work per agency client

SaaS reporting platform for Yandex Direct

Own product for digital agencies

Challenge:: Some agencies and freelancers still spend hours on manual reporting for every client: exports from Yandex Direct, cross-checking with Metrika, cost-per-conversion math, formatting and delivery. They need a service that runs all their clients by itself — from sync to a ready report and alerts.
Built:: Built a full-cycle multi-tenant SaaS: team workspaces with roles, Yandex OAuth (tokens encrypted, auto-refreshed), background sync of campaigns, spend, account balances and Metrika data via Horizon queues. Conversions are weighted and mapped to Metrika goals. Analytics is two-layered: deterministic rules catch anomalies statistically (z-score against each client's own baseline), with an LLM assessment layered on top producing per-client takeaways. Reports are delivered to end clients via public share links; alerts and digests arrive in Telegram. Under the hood: billing with trials and auto-charging, a super-admin panel, onboarding, PWA. Quality is enforced by CI — tests, PHPStan, Pint on PHP 8.3 and 8.4.

Complete SaaS cycle: sign-up → trial → billing with auto-charge, no manual administration
Two-layer insights: statistical rules plus LLM analysis for every client
Reports via public share links, anomalies and balances in Telegram; in daily production use

LaravelFilamentLivewireHorizonOAuth 2.0Yandex Direct APIMetrika APILLM insightsTelegram Bot APIPWACI

200,000+offers from 1,400 suppliers in a single catalog

B2B platform for a concrete products wholesaler

ZhBI Rossii

Challenge:: A wholesale supplier of reinforced concrete products needed one system instead of scattered spreadsheets: a catalog of hundreds of thousands of items from ~1,500 suppliers, quote calculation, logistics, price control — built around how their managers actually work, not a generic CRM template.
Built:: Built a B2B platform around a three-level catalog (product → variant → offer) with a synonym system for matching. Supplier price lists are imported with LLM-powered parsing; product attributes are enriched via AI web search. Price anomalies are caught automatically with median-based statistics. The quoting module calculates offers across 1,400 suppliers and generates proposals and invoices; manager-adjusted prices flow back into the catalog through an approval queue. Plus logistics with PDF powers of attorney, counterparty due-diligence via registry data (financials, litigation), a Telegram bot and search analytics. The platform shares its database with the company's public B2B portal.

200,000+ offers from 1,400 suppliers — a catalog with variants, synonyms and instant search
AI handles imports and catalog enrichment; the system finds price anomalies on its own
Quote → proposal → invoice → logistics: the full deal cycle in one system, new modules shipping weekly

LaravelFilamentLivewireMeilisearchLLM importDockerTelegram Bot APIDomPDF

24/7runs unattended in production

IoT weather station: ESP32, dashboard and Android app

Personal project

Challenge:: A full-cycle personal testbed: from hardware to a mobile app. ESP32 sensors measure temperature and humidity, and the data has to turn into convenient analytics on any device.
Built:: Built the whole stack myself: the ESP32 firmware sends readings to the server, the backend stores them in MySQL, an interactive ECharts dashboard renders charts from a day to a year with algorithmic smoothing of sensor outliers. A Telegram bot delivers forecasts, and the system also compares forecasts against actuals, tracking accuracy by time of day. Plus a native Android app.

Runs 24/7 unattended, with years of observation history
Analytics: records, year-over-year comparison, forecast accuracy by time of day
Full cycle in one pair of hands: hardware → backend → dashboard → mobile app

ESP32 / C++PHPMySQLEChartsTelegram Bot APIAndroid

2 → 25–30leads per month — same ad spend

Wholesale steel pipes website

StalEnergo-Ural

Challenge:: The old WordPress site barely generated leads: about 2 form submissions per month despite active ads (~10 Yandex Direct campaigns).
Built:: A completely new site: fast catalog with filters, admin panel for price lists and content, SEO structure, forms and analytics.

25–30 leads per month instead of 2 — same ad campaigns
A lead almost every business day, more phone calls too
The client updates content and prices via the admin panel

LaravelFilamentTailwind CSSAlpine.jsSQLiteSEO facetsVisit site →

How I work

AI as a force multiplier

Claude Code and Codex give me the throughput of a small team. Judgment, architecture and accountability stay with me.

Engineering discipline

Tests, static analysis, code style checks and CI on every project — speed without a quality tax.

Working software over slides

First deployable version in days. Iterations you can click, not status meetings.

Measurable outcomes

Leads, sync times, uptime — every project has before/after numbers, not adjectives.

Background

Thirteen years on the marketing side of the web taught me what software is actually for: traffic, conversions, decisions. Then AI-assisted development made it possible to build at the speed I always wanted — so I went all-in and shipped four production systems in a year.

2012
Started in digital advertising; grew to 70+ Yandex Direct projects as an independent specialist
2013–2018
Five years of SEO in parallel: site structure, semantics, technical optimization
Now
Full pivot to AI-assisted development: CRM, SaaS, e-commerce rebuild, IoT + Android — all in production

Let's talk

Hiring, contracting or just curious how AI-assisted development works in practice — I'll reply personally.

Message me on Telegramor by email [email protected]

github.com/aa-gavrilov