Full-Stack Developer · AI Engineer · Seoul, South Korea
I build full-stack products and then put real AI inside them — semantic and vector search, streaming assistants grounded in live application data, and LLM endpoints with hard cost guardrails that degrade gracefully instead of falling over. Four years across e-commerce, healthcare, real estate, retail inventory and consulting SaaS, on a TypeScript / NestJS / Next.js core with Java Spring Boot behind it.
I own things end to end: schema, API, UI, Docker image, Nginx config, and the deploy that puts it in front of users.
ROLE="Full-Stack Developer / AI Engineer"
EXP="~4 years (since May 2022)"
DOMAIN="E-commerce · Healthcare AI · Real Estate · Retail Inventory · Consulting SaaS"
STACK="TypeScript · Next.js · NestJS · Express · Spring Boot · GraphQL · MongoDB · PostgreSQL"
AI="OpenAI · Claude (Skills/MCP/Agents) · Gemini · LangChain · Groq · Ollama · vector search"
OPEN_TO="Full-time in South Korea — onsite / hybrid / remote · freelance contracts"Languages
Frontend
Backend & API
Databases
Cloud & DevOps
CV & Tooling
| Domain | Proficiency | Details |
|---|---|---|
| Backend & API architecture |  |
NestJS + Apollo GraphQL code-first, Express 5, Spring Boot. 40+ REST endpoints on one platform, 32 GraphQL operations across 11 modules on another. JWT with refresh-token rotation, up to 4-level RBAC. |
| AI product engineering |  |
MongoDB Atlas vector search, SSE streaming assistants with tool calling, multi-model fallback chains across OpenAI / Gemini / Groq / Ollama, Redis-tracked daily spend caps. |
| Frontend & product UI | |
React 19, Next.js App Router, Tailwind v4, shadcn/ui, Framer Motion. Role-based dashboards at 15+ pages; 40% page-load reduction via SSR optimization. |
| Data modelling & query perf |  |
MongoDB compound indexes and aggregation pipelines (60% query-time cut), $facet pagination, PostgreSQL, MySQL/MariaDB, Prisma / TypeORM / Mongoose. |
| DevOps & deployment |  |
Docker multi-stage builds, Nginx reverse proxy, PM2 zero-downtime releases, GitHub Actions, AWS EC2/S3 with automated backups, Vercel and Render. |
| Realtime systems | |
Socket.io notifications and messaging with read receipts and typing indicators, raw WebSocket gateways, Twilio DataTrack control channels. |
Shopy — AI e-commerce with semantic search and a streaming shopping assistant
Semantic product discovery over a 560-product catalog, plus a multi-turn assistant that streams answers grounded in live catalog data instead of hallucinating inventory.
| Stack | Next.js 16 · Express 5 · MongoDB Atlas vector search · OpenAI gpt-4o · Stripe · Upstash Redis |
| Scale | 560-product catalog indexed for vector search; multi-turn conversation state |
| Impact | Vector search pipeline, SSE streaming assistant with tool calling, recommendation fallback chain, Redis-tracked daily AI budget cap so spend degrades gracefully instead of breaking |
| Role | Sole developer |
Livit Platform — real estate marketplace on code-first GraphQL
Listings, agent directory and real-time messaging for a property marketplace, built on a modular NestJS + Apollo backend.
| Stack | NestJS 11 · Apollo GraphQL (code-first) · MongoDB · Next.js 16 · Google Maps |
| Scale | 32 GraphQL operations across 11 modules |
| Impact | $facet aggregation pagination, WebSocket messaging, full schema and module architecture |
| Role | Sole developer |
TwilioMeet — privacy-first video interview platform
Video interviews with a waiting room and explicit admit flow, live interviewer notes and candidate rating — no candidate joins the room before they are let in.
| Stack | React 19 · Vite · twilio-video · Node/Express 5 · TypeScript · Vitest |
| Scale | 8 test suites; single DataTrack carrying both chat and control messages |
| Impact | Stateless role model encoded directly in the Twilio identity, server-side token minting, admit/deny gate |
| Role | Sole developer |
JQ-Auto (UpCar) — inspection sheets and multi-page PDF generation for parts exporters
Freelance tool letting auto-parts exporters assemble inspection sheets and export them as multi-page PDFs that match the on-screen preview exactly.
| Stack | React 19 · Express 5 · Puppeteer · MongoDB · Cloudinary |
| Scale | Handles 500-row documents |
| Impact | WYSIWYG preview sharing one HTML/CSS source with the PDF renderer, singleton Puppeteer instance to keep memory flat |
| Role | Sole developer, freelance |
Avicenna — multilingual AI symptom analysis for foreigners in Korea
EN / KO / UZ symptom analysis aimed at foreign residents navigating Korean healthcare, with exportable medical history.
| Stack | React 19 · Firebase · Gemini API · Vercel Functions |
| Scale | 3 languages, multi-model fallback chain |
| Impact | Fallback chain across models so the assistant stays available, medical history with PDF/JSON export |
| Role | Sole developer |
AI Customer Support Automation — ticket triage and routing
Classifies inbound support tickets, scores priority and routes each one to the right team automatically.
| Stack | n8n · OpenAI · Slack · Notion |
| Scale | End-to-end inbound triage workflow |
| Impact | Classification, priority scoring and team routing without manual triage |
| Role | Sole builder |
Jan 2024 – Dec 2025 · Remote
- Built a client management platform serving consulting clients across 4 countries
- Shipped the Next.js / TypeScript dashboard — 15+ pages with role-based views
- Built the NestJS + PostgreSQL backend behind it — 40+ REST endpoints
- Implemented Socket.io real-time notifications and messaging with read receipts and typing indicators, secured by JWT + refresh-token rotation and 4-level RBAC
- Owned deployment end to end — Docker multi-stage builds, Nginx reverse proxy, PM2 zero-downtime releases — and cut page load time 40% via Next.js SSR optimization
May 2022 – Nov 2023 · Remote
- Built a Java / Spring Boot product catalog and inventory API — 30+ endpoints for a multi-store retailer
- Modelled and served 2,000+ SKUs across multiple store locations
- Cut query time 60% with MongoDB compound indexes and aggregation pipelines
- Built multi-store inventory sync with conflict resolution and audit logging
- Set up Spring Security with 3-level RBAC, AWS EC2/S3 infrastructure with automated backups, and Swagger/OpenAPI docs for the frontend team
| Result | Where | How |
|---|---|---|
| 60% faster queries | ISHONCH Electronics | MongoDB compound indexes + aggregation pipelines |
| 40% faster page loads | Silk Road Professionals | Next.js SSR optimization |
| 4 countries served | Silk Road Professionals | Client management platform |
| 2,000+ SKUs managed | ISHONCH Electronics | Multi-store inventory sync with conflict resolution |
| 560-product AI catalog | Shopy | MongoDB Atlas vector search pipeline |
| 32 GraphQL ops / 11 modules | Livit Platform | NestJS + Apollo code-first schema |
| 500-row PDF documents | JQ-Auto | Singleton Puppeteer render pipeline |
| 8 test suites | TwilioMeet | Vitest coverage on realtime logic |
| Certification | Issuer | Date |
|---|---|---|
| The Complete Full-Stack Web Development Bootcamp | Udemy — Angela Yu | Sep 2025 |
| Connect and Protect: Networks and Network Security | Google / Coursera | Jan 2024 |
| Junior Penetration Tester | TryHackMe | Nov 2024 |
Dong-A University — Bachelor of Business Administration, International Business Management
2022 – 2025 · Busan, South Korea
learning:
- AI agent engineering — Claude Skills, MCP servers, multi-agent orchestration
- Backend architecture at scale — NestJS module design, GraphQL federation patterns
- Java Spring Boot — deeper into Spring Security and JPA internals
building:
- Shopy — hardening the vector search + streaming assistant toward production
- Avicenna — multilingual AI symptom analysis, production readiness
- OpenClaw — agent tooling experiments
exploring:
- MongoDB Atlas vector search tuning and embedding cost control
- Local inference with Ollama and Groq for latency-sensitive AI endpoints
- n8n workflow automation as glue between LLMs and business tools
open_to:
roles: [Full-Stack Developer, Backend Developer, AI Engineer]
location: South Korea — onsite / hybrid / remote
contract: Freelance projects welcome
visa: D-10 job seeker
contact: alexnabiyev5@gmail.comOpen to full-time roles in South Korea — onsite, hybrid or remote — and to freelance contracts.