// problem solver · full-stack developer

I turn complex problems into working solutions.

Algorithms, automation scripts, APIs and AI systems - from a muscle-coverage engine that recalculates in real time to WordPress platforms serving live search traffic. If it needs real logic, I'm interested.

What I do See the proof

About

An experienced full-stack developer who solves the hard part.

I'm a full-stack developer who thinks in terms of what a problem actually requires - the algorithm, the data model, the integration - not just the interface on top of it. My range runs from PHP and WordPress to Node, React, and Next.js, with AI APIs wired in wherever they solve a real problem.

My process is the same regardless of stack: pull the problem apart, find the logic that actually solves it, ship it, and keep it working once it's live. That shows up as a muscle-coverage algorithm, a streaming filter bot, a multi-client WordPress platform, and AI pipelines running unattended in production.

Haifa District, Israel
Coding professionally since 2021
Lead Full-Stack Developer @ Hitsearch
Hebrew · English · Russian

What I do

Four things I'm actually good at.

4 core skills

Complex solutions, algorithms & scripts

When the problem is the logic itself.

Muscle-activation math that recalculates a training plan in real time, a streaming filter bot that separates signal from noise in a live feed - the kind of work where the hard part is the algorithm, not the UI.

Proves it: Full Coverage · Twitter Stream Filter Bot · Lab Admin - Technion EE Labs

Complex WordPress themes & plugins

Major experience

Not page-builder WordPress - platform WordPress.

One reusable theme powering many branded client sites, configured entirely from the admin - no per-client code. A companion plugin that generates hundreds of SEO-structured pages from live sports data. My deepest, most repeated area of work.

Proves it: SEO Blog Platform

AI APIs & AI systems

LLMs wired into real pipelines, not demos.

Streaming chat over Claude and OpenAI inside a clean, domain-driven backend. A queue-driven pipeline that ranks products, generates copy with an LLM, and composes branded images - unattended, at scale.

Proves it: AI Content Creator · Affiliate Content Pipeline

APIs & backend systems

The part users never see and always feel.

Payment integrations with real success/failure/cancel flows, job queues, dockerized multi-container systems, clean architecture that survives contact with production.

Proves it: Akko Ticketing Kiosk · Lab Admin - Technion EE Labs · AI Content Creator · Affiliate Content Pipeline

Also working with

PHP
Next.js
React
Laravel
Node.js
TypeScript
Vue 3
MongoDB
MySQL
Docker

Selected work

Proof, not promises.

7 projects

Own productLive

Full Coverage

My own SaaS, live in production. Every exercise variant carries a muscle-activation vector, and the app sums activation across your session to project weekly per-muscle coverage on a body heatmap. A training mode walks you through a session, remembers where you left off, and greedily suggests swaps to patch any muscle coming up short. Bilingual EN/RU.

Node.js
Canvas
Multi-user
i18n

Case study

Problem

"Which muscles did my week actually cover?" is a math problem that most trackers dodge by just logging sets.

Approach

Gave every exercise variant a muscle-activation vector instead of a single body-part tag.
Summed activation across a session to project a live weekly per-muscle coverage percentage.
Rendered coverage on a body heatmap that updates the instant a set is logged.
Built a training mode that remembers where you left off and greedily suggests a swap when a muscle is under-covered.
Kept the stack intentionally minimal - plain Node `http`, no frontend framework - by design.

Hard parts

Tuning activation weights until the numbers matched real training intuition, not just a formula.
Making the recalculation feel instant on every swap, not a page reload.
Shipping it bilingual (EN/RU) without forking the logic.

Outcome

Live in production at fullcoverage.app.

fullcoverage.app ↗

ScriptPrivate code

Twitter Stream Filter Bot

A streaming bot that filters a live Twitter firehose down to relevant stock-related tweets and acts on them automatically.

Node.js
Twitter API
Streaming
Email

Case study

Problem

A live tweet stream is mostly noise - the useful fraction has to be found in real time, not after the fact.

Approach

Connected to the Twitter streaming API with custom stream rules to narrow the firehose to stock-related tweets.
Wrote a custom filter to catch and drop reposts before they reached the output stage.
Wired matches to two outputs: an email alert and an automatic retweet.

Hard parts

Keeping a long-running stream connection alive and recovering cleanly from drops.
Deduplicating reposts without a heavyweight matching approach.

AIPrivate code

AI Content Creator

Chat-based content tool over Claude & OpenAI, built as a clean, domain-driven TypeScript system.

Node.js
Express
TypeScript
DDD
React
MongoDB

Case study

Problem

Chat-based content generation is easy to prototype and easy to let rot into prompt spaghetti once it has to support more than one AI provider.

Approach

Structured the backend as Clean / Hexagonal Architecture with Domain-Driven Design: rich domain entities, repository and provider patterns.
Abstracted Claude and OpenAI behind a shared provider interface with dependency injection.
Validated every boundary with Zod so bad input fails fast, not deep in the domain.
Streamed responses through to a React + Vite + Zustand feature-based frontend in real time.

Hard parts

Keeping streaming responses clean through multiple architectural layers instead of one giant handler.
Designing a provider abstraction that would survive future API changes on either vendor.

AIPrivate code

Affiliate Content Pipeline

Queue-driven system that generates branded, platform-ready affiliate posts with AI copy and composed images.

Next.js 15
React 19
BullMQ
Redis
Groq
MongoDB

Case study

Problem

Manually writing and designing affiliate posts for multiple platforms does not scale past a handful of products a day.

Approach

Built a ranking step to surface which AliExpress products are worth posting.
Generated platform-tuned copy with Groq LLaMA 3.3 70B for Instagram, Facebook, and TikTok separately.
Composed branded post images on the fly with Canvas instead of static templates.
Ran generation as BullMQ + Redis background jobs with a live dashboard to watch and retry them.

Hard parts

Keeping an unattended job queue reliable - retries, failures, and backpressure all had to be handled, not assumed away.
Making Canvas-composed images look designed, not generated.

WordPressClient work

SEO Blog Platform

A reusable multi-client WordPress SEO platform - one theme, many branded sites - plus a World Cup 2026 data plugin.

WordPress
PHP 8
Schema.org
RTL
Polylang

Case study

Problem

Every new affiliate client meant another WordPress install - the old way meant another set of one-off template edits per client too.

Approach

Built one reusable theme where brand colors, logos, SEO settings, and an on-page section builder are all configured from the admin - zero per-client code.
Added automatic RTL support so Hebrew and other RTL-language clients work out of the box.
Shipped a companion World Cup 2026 plugin that auto-generates 32 team, 8 group, and 64 match pages from live football data.
Gave every generated page full Schema.org, Open Graph, and JSON-LD markup plus configurable affiliate CTAs.

Hard parts

Making "everything configured in admin" actually hold up across real, different client sites - not just the demo.
Keeping structured SEO markup correct at generated-page scale (100+ pages from one data feed).

PaymentsClient work

Akko Ticketing Kiosk

Self-service ticketing kiosk for Akko tourism - museums and guided tours - with card payments.

Laravel
PHP
Pelecard
MySQL

Case study

Problem

An unattended kiosk selling real tickets for real money has nobody standing by to help when a payment fails.

Approach

Built per-terminal configuration so each kiosk can be tuned to its location and ticket set.
Modeled ticket types and visitor types to match how Akko actually sells tours.
Added digital waiver signing directly on the kiosk.
Integrated Pelecard payments end to end, with explicit success, failure, and cancel flows.
Shipped Hebrew + English, run through a custom admin.

Hard parts

Covering every payment edge case - declined cards, timeouts, cancellations - with no human in the loop to recover manually.

SystemsClient work

Lab Admin - Technion EE Labs

A lab-management system for Technion electrical-engineering labs, deployed as a dockerized multi-container stack.

React
Node.js
PostgreSQL
Docker
nginx

Case study

Problem

A university lab needs real role separation between students and teachers, and it needs to keep running without a dedicated ops team.

Approach

Modeled the domain around distinct student and teacher roles and permissions.
Deployed as a five-container Docker stack: React client, Node API, nginx, PostgreSQL, pgAdmin.
Followed 12-factor principles, keeping all production config out of code.
Orchestrated the whole stack with docker-compose for a repeatable deploy.

Hard parts

Getting role and permission modeling right for a system with real students and real teachers using it daily.
Running production infrastructure without a platform team behind it.

Some projects are client or commercial work - described in full, but the code stays private. More on GitHub ↗.

Experience

Where this was built.

Lead Full-Stack Developer · Hitsearch

End-to-end technical ownership: architecture, infrastructure, and every production system, from AI tooling to the servers underneath.

Architect and run the company's entire technical infrastructure - servers, deployments, and every production system.
Built an AI content-creation tool over Claude and OpenAI on a clean, domain-driven backend.
Designed and run a reusable WordPress theme powering multiple branded client SEO sites from one codebase.
Shipped a World Cup 2026 plugin that auto-generates hundreds of SEO-structured pages from live sports data.

PHP
WordPress
Next.js
Node.js
TypeScript
MongoDB

Built here

Contact

Have something to build?

Open to interesting projects and collaborations. The fastest way to reach me is email.

mikezinodev@gmail.com

GitHub ↗LinkedIn ↗