M. Kodirov

Mukhammadali Kodirov

Backend engineer in Kobe, Japan. I build Node.js, Go and PostgreSQL systems on AWS, cut what they cost to run, and connect LLMs to them with fallbacks that hold up.

−57% RDS cost−30% dev and staging cost0 card numbers handled by staff

Open to remote Senior Backend and Backend + AI roles

Read the case studiesGet in touch
CustomersALBEC2RDSS3SESPaymentsVPC

Selected work

Each project names the problem, what I changed, and what it was worth. Client names are left out.

PostgreSQL 13 to 16 on AWS RDS

A production RDS database was still on PostgreSQL 13.20. Standard support had ended, so AWS was adding Extended Support surcharges to every monthly bill.

  • Spotted the surcharge in billing and traced it to the engine version.
  • Took RDS snapshots and application-state backups, and moved every connection to enforced SSL.
  • Upgraded to PostgreSQL 16.3 in a scheduled maintenance window, then smoke-tested the live user flows.
Read the write-up
−57%monthly cost

Infrastructure cost went from $700 to $300 a month, and every database connection now uses SSL.

AWS RDSPostgreSQL 16Node.js

Tokenized card payments on AWS

On an AWS-hosted electricity platform, staff updated customer cards by hand from CSV files. They saw raw card numbers, and each update took about half a day.

  • Led the integration as project manager, using the provider's hosted payment page so card data never touches our servers.
  • Gave customers self-service card updates, with the API storing only payment tokens.
  • Scoped API permissions so only the owning customer can create or read a payment.
0cards seen by staff

A half-day manual task became instant self-service, and staff no longer handle card numbers at all.

GoAWS EC2ALBPaygentAuth0

Dev and staging that shut down at night

Development and staging ran on cloud databases and an RDS Proxy around the clock, for a team that works Tokyo hours.

  • Replaced the dev RDS Proxy with containerized local databases.
  • Added a scheduled Lambda that stops staging and dev on nights and weekends.
  • Kept every environment on the same schema with Prisma migrations.
Read the write-up
−30%per month

Lower monthly cloud cost, and near-zero database latency for the dev team.

AWS LambdaEventBridgeDockerPrisma

AutoReach Pro: vision AI for Japanese business forms

A Chrome extension that reads Japanese company contact forms and fills them in. Layouts vary so much that no single model handles every page.

  • Built a vision pipeline that starts with Gemini 2.5 Flash-Lite and falls back to Gemini 2.5 Flash, then Claude Haiku.
  • Routes DeepSeek to cleaned DOM only, since it has no image input.
  • Runs analysis jobs through BullMQ so the extension never waits on a slow model.
In development

The cheapest model handles most pages. Stronger models take over only when it fails.

GeminiClaude APIBullMQChrome Extension

Experience

Five years of shipping web backends, most recently leading client projects in Japan.

  1. Software Engineer

    Quin-Que Inc., Kobe

    I run client projects end to end, from requirements to release, across web, mobile and AWS infrastructure.

    • Cut a production database bill by 57% ($700 to $300 a month) with an RDS PostgreSQL upgrade.
    • Replaced staff-handled card updates with tokenized self-service payments, turning a half-day task into an instant one.
    • Found an idle dev database billing $178.56 a month in AWS Extended Support, snapshotted and deleted it, and fixed the billing alarms.
    • Develop and maintain an internal HR platform for attendance, leave and overtime approvals: 124 API routes on Next.js 16, Prisma and MySQL.
    • Moved a WordPress site off AWS EC2 to a new server, moving files with rsync and the database with MariaDB.
    Internal HR platform: attendance, leave, overtime, hiring examsNext.js 16, Prisma, MySQL, Docker
    Tokenized card payments for an electricity platformGo, AWS, Paygent
    Sales automation: finds target companies, drafts outreach with AI, submits contact formsNext.js, Supabase, Vercel
    WordPress migration off AWS EC2EC2, rsync, MariaDB
    Delivery management mobile app for a trading companyFlutter, Firebase
    LINE bot for travel-distance tracking and expense claimsNestJS, Next.js, MySQL
    Browser-based glass film simulator with product adminNext.js, Prisma, MySQL

  2. CTO

    Yetti.jp, Kobe

    Technical lead for a small web, mobile and AI studio serving Japanese and international clients. I'm building AutoReach Pro, the vision-AI form-filling extension above.

  3. Backend Developer

    Navana Inc.

    • Designed and built the backend for an early-stage job marketplace connecting job-seekers with employers across Uzbekistan.
    • One set of APIs served both the website and the mobile app, reaching 800+ users across both.

  4. Full-stack Developer, part-time

    Algoritm Learning Center, Namangan

    • Replaced paper and Excel point tracking with a MERN web app and a Telegram bot for teachers.
    • Weekly point collection dropped from 3 days to half a day (−83%), and students saw their standings within seconds.
    • Every teacher and student at the center used it.

  5. Customer Service Manager

    Phobos, Japan

    Customer-facing work in Japanese before I moved into software.

Writing

Notes on AWS cost, databases and running a small engineering team. All posts

Hiring for a remote backend role?

Email is the fastest way to reach me. I work in English, Japanese and Uzbek.

Emailkodirov8788@gmail.com
LinkedInin/mukhammadalikodirov
GitHubkodirov8788
Based inKobe, Hyogo, Japan