Senior Flutter Developer · Go · LLM Integrations

Anton Kozyrev

Flutter developer with 4.5+ years of commercial mobile development. For the last three years I've been the sole mobile developer behind three products: architecture, stack, API contracts, releases. I build my own products end to end — Go backend, web, LLM integration, deployment.

  • Kaliningrad, Russia · Remote
  • Open to projects and full-time offers
  • 4.5+yrs Flutter
  • 10mobile apps
  • 3products as sole mobile dev
agent · session
  1. task: add a monthly budget guard to the LLM client
  2. plan: check worst-case cost before each call → fail fast → test
  3. ~ internal/llm/client.go
  4. + internal/llm/budget.go
  5. + internal/llm/budget_test.go
  6. 3 files to change. Approve?
  7. go test ./... — ok
  8. diff reviewed · secret scan clean

01Selected work

Flutter · 4.5+ years

Commercial mobile apps

Apps for contractors of a staffing service — GSR Rabota and Contractor Portal: onboarding, documents with SMS signing, shift booking, payouts and tax receipts. Sole mobile developer on the product team; the code is under NDA, the apps are in the stores — see Apps below.

  • 10 mobile apps — from greenfield launches to taking over code from other teams
  • GSR Rabota — built from scratch: architecture, stack, API contracts with the backend, releases
  • Contractor Portal — migrated from native Android (Java) to Flutter with full feature parity, shipped on iOS and Android, new design; now maintaining and developing version 3.x
  • Feature-first Clean Architecture, Riverpod for state, API contracts co-designed with the backend team, real-time chat on Socket.IO, FCM push notifications, Crashlytics, Remote Config
  • Key logic covered with unit and integration tests; legacy refactoring without stopping feature work

Flutter · Dart · Riverpod · GoRouter · Dio · Socket.IO · Firebase · AppMetrica · Drift/SQLite

Go · LLM · content pipeline

sitefarm — data-driven site engine

A Go engine that turns open data into a content site: collect data (NOAA, PRISM) → write text with an LLM → fact-check with a second model → build a static site.

  • Numbers come only from the data, never from the model; a second model checks every text against the facts
  • Deterministic checks run before the reviewer model (dates, 10/50/90% probabilities, made-up people); they cut the main error, an average date presented as the frost date, from 37% to 0%
  • The final prompt passed the fact check on 90% of pages (27 of 30) in a trial run; a failed page is never published: one automatic retry, then a manual review queue
  • Budget guard: the model is not called if the worst case would exceed the monthly limit
  • Models benchmarked on the real task: a reasoning model spent its whole token limit on reasoning and returned nothing, so I dropped it
  • About 4–5 RUB (≈5 US cents) per page, regenerations included

Go · Yandex AI Studio (gpt-oss-120b) · SQLite

Hugo · static site · US

localfrostdates.com — frost dates for US gardeners

The first site built with sitefarm: last spring and first fall frost dates for US weather stations, with odds instead of a single date.

  • Live since 2026-10-29: 2,088 NOAA stations, 1991–2020 climate normals
  • Search with suggestions across ~16,800 US cities and ZIP codes, no third-party services; an accessible ARIA combobox
  • “Use my location” finds the nearest stations: distance is computed in the browser, coordinates never leave the device
  • Every station page has a year-at-a-glance chart, 10/50/90% frost probabilities and a map of nearby stations
  • No cookies or third-party scripts; CSP with script hashes

Hugo · Cloudflare Pages

Built for a US audience; may not load from Russia without a VPN.

Go · PostgreSQL · PWA · LLM

Terroir — personal wine assistant

An assistant with memory: cellar, taste profile, wine matched to your own recipes. Retrieval-grounded answers: the code picks the data, the model only explains it.

  • Answers are validated: the model cannot reference a bottle that is not in the cellar
  • Monthly token budget and a fallback when the LLM fails
  • 90+ tests, CI on GitHub Actions

Go · PostgreSQL/SQLite · PWA · OpenAI-compatible LLM client · site: Astro, EN/RU

B2B SaaS · in development

TapReply — replies to customer reviews

AI-drafted replies to customer reviews across several platforms, for restaurants and other venues, in local languages. A Go core with platform adapters: a new review platform is a new adapter, not a rewrite.

  • Go core + platform adapters
  • Flutter / PWA client
  • PostgreSQL

Go · Flutter · PWA · PostgreSQL · site: Astro, EN/RU

02Apps I built

The first two are from my current job. I built the rest alongside it — separate repos, clear scopes, every app shipped to the stores.

Current job
GSR Rabota — screenshot 1
GSR Rabota — screenshot 2

GSR Rabota

2026 · Flutter

New app for contractors of a staffing service: sign-up and documents, SMS signing, finding tasks and booking shifts, wallet and payouts, tax receipts for the self-employed, real-time support chat.

Sole mobile developer, built from scratch: architecture, stack, API contracts with the backend, releases.

Flutter · Riverpod · GoRouter · Dio · Socket.IO · Firebase · AppMetrica · Drift/SQLite

Current job
Contractor Portal — screenshot 1
Contractor Portal — screenshot 2

Contractor Portal

2025 · Flutter

Contractor app of the same service: shift self-booking, documents with e-signing and receipts, profile and bank cards. In the stores as «Личный кабинет исполнителя».

Migrated native Android (Java) to Flutter and shipped it on iOS and Android; now maintaining and developing version 3.x.

Zhabby — screenshot 1
Zhabby — screenshot 2

Zhabby

2025 · Flutter

Short-video social network that rewards activity: personal feed, video upload, points for views, likes and comments, rewards exchange.

Second developer on the team: built every screen from Figma designs, wired up API calls and state management in several sections. Published on the App Store and Google Play.

Mealhub — screenshot 1
Mealhub — screenshot 2

Mealhub

2022–2023 · Flutter

Restaurant search on a map, table booking and food ordering.

Second developer: responsive UI from Figma, Google Maps API, state management, REST API integration, bug fixing and feature work.

GreenFeed — screenshot 1
GreenFeed — screenshot 2

GreenFeed

2022–2023 · Flutter

Animal feed catalog and feed-ration calculator for farmers.

Second developer: UI from Figma, state management, REST API integration, bug fixing and feature work.

Jobline

2023–2024 · Flutter

Platform for working with external contractors: contractor onboarding, documents and receipts, payouts with tax withholding for self-employed contractors.

Built the mobile app from scratch.

Q-Working

2023–2024 · Flutter

Service that finds contractors, handles deals and pays them.

Built the contractor app from scratch: onboarding, profile and services, orders and payouts.

03What I can help with

Flutter features and releases

New screens and flows from Figma to the App Store and Google Play: state, API integration, push notifications, crash reporting.

Taking over Flutter code

Picking up someone else's codebase, refactoring legacy code without stopping feature work, covering key logic with tests.

Go backend and MVP API

REST API in Go with PostgreSQL and CI on GitHub Actions — enough to launch a product and validate the idea.

LLM features with guardrails

Budget limits, fact-checking and fallbacks: the model does not invent data and does not burn the budget.

A one-off project, a long-term contract or a full-time role — tell me what you need, and I'll come back with a clear plan and an estimate. Get in touch →

Discuss a project →

04How I work with AI agents

Agents take on the volume of work; responsibility for the result stays with me.

Agent toolchain
  • Veaidaily driver
  • Cursoralso used
  • GitHub Copilotalso used
  • Claude Codesame workflow
  • OpenAI Codexsame workflow
  1. 01

    Agent in the IDE, plan first

    My daily driver is Veai (AI coding agent for JetBrains IDEs), in IntelliJ. It reads the project, edits code, runs tests and commands. Plan and decomposition come first, code second; the agent keeps project memory, my own skills and rules. I have also worked with Cursor and GitHub Copilot; Claude Code and OpenAI Codex follow the same agentic workflow — plan → diff → tests — so switching between them is quick.

  2. 02

    Parallel agents, review before merge

    In one session I ran six agents in parallel, each in its own repo (Astro/Hugo sites, PWA clients, Go backend), ~2 hours of staged work with acceptance criteria. Then I review: tests, builds, commit authorship, secret scan. Within one repo, parallel agents get separate git worktrees. Nothing lands in main without this review.

  3. 03

    LLMs inside the product, with guardrails

    In sitefarm, texts are generated through Yandex AI Studio and checked by a second model; in Terroir, the code selects the data and every answer is validated against the database. Both have hard monthly budgets.

05Stack

Each skill links to where I've used it — not a tag cloud.

Mobile
Flutter, Dart, Riverpod, GoRouter, Dio, Socket.IO, Firebase (FCM, Crashlytics, Remote Config), Drift/SQLiteCommercial apps · TapReply
Backend
Go, REST API, PostgreSQL, SQLite, CI/CD (GitHub Actions)terroir-monorepo · sitefarm-showcase
LLM
OpenAI-compatible APIs, Yandex AI Studio, second-model fact-checking, budget limits, fallbackssitefarm · Terroir
Web & infra
Astro, Hugo, PWA, Cloudflare (Pages, DNS), Linux VPSlocalfrostdates.com · this site
Architecture & quality
Clean Architecture, feature-first, unit and integration tests, legacy codeCommercial apps · Terroir tests
Anton Kozyrev

About

Senior Flutter developer from Kaliningrad. Before moving into development, I spent five years managing projects and led a six-person department — so I start from the business problem, not the code. Today: mobile apps in Flutter, Go backends and LLM features with guardrails.

06Contact

Tell me what you need — a project, a long-term contract or a full-time role.