👋 Welcome! A tech blog where a backend developer shares experiences in Backend, Infrastructure, Troubleshooting, and more ✨

Contact

jji042842@gmail.com · GitHub · Blog

Profile

I’m Jeongil Jung, a backend developer with 3 years of experience, always asking “What improvement does our service need most right now?”

From API performance optimization to MSA migration, K8s infrastructure setup and operations, and team development culture— if needed, I proposed it and built it with my team.

I completed our MSA migration with zero downtime, and reduced incident detection time from 1 hour to 1 minute.

Core Strengths: MSA Migration Design & Implementation · K8s Infrastructure · Query & API Performance Optimization · Monitoring System

Skills

SkillExperience (Years)Description
Java3.0Backend development in MSA, Batch environments
Kotlin3.0Async processing with Coroutines, primary language in MSA
Spring Boot3.0REST API development with Spring Security, Actuator
Spring Batch3.0Large-scale data processing, Chunk/Partitioning optimization
Spring Cloud2.5MSA with Gateway, Eureka
MySQL3.0Certified - SQL Developer | DB design, query tuning, Replication
Kubernetes2.5Certified - CKA | GKE infrastructure, Helm, ArgoCD automation
Linux4.0Certified - Linux Master Level-II, AWS CCP | Server operations

Work Experience

Bootalk Inc.

AI-based real estate brokerage platform (210K+ users, 50K+ broker matches)
2025.03 ~ Present ( 1 years 0 months ) | Backend Engineer (4-member dev team)

Tech Stack: Kotlin Spring Cloud Kubernetes ArgoCD Grafana LGTM Stack Redis MySQL

  • Monolith → MSA Migration Architecture Design & Implementation

    Decoupled tightly coupled domains, zero-downtime transition completed, common code changes 16 PRs → 1 PR
See Details
  • Overview: Zero-downtime migration from Java Servlet monolith to Kotlin/Spring Cloud MSA
  • Duration: 2025.04 ~ 2025.11 | 3-member team, designed and implemented 5 of 18 services (Property Listings, Broker-User Matching, Transaction Prices, User, Notification)
  • [Problem]
    • Tightly coupled core domains (listings, broker-user matching) blocked gradual migration strategy
    • Common code scattered across 16 repos → 16 PRs required for same change, version mismatch
  • [Key Contributions]
    • After initial gradual migration attempt failed, designed partial big-bang strategy – migrated tightly coupled domains simultaneously
    • Dual Write + validation batch for data consistency during transition
    • Hexagonal Architecture + CQRS to remove technical dependencies from domain logic
    • Multi-module structure design – after initial Common bloat issue, limited to Type/Enum/Util only, separated infra code to dedicated modules
  • [Results]
    • Zero-downtime MSA migration completed, Hexagonal Architecture enabled independent domain deployment
    • Common code changes: 16 PRs → 1 PR, Build time: 27min → 8min
  • Tech: Kotlin, Spring Cloud, Feign Client, CQRS, Hexagonal Architecture, AWS SQS

  • Kubernetes Infrastructure & Monitoring System

    From no monitoring, incident detection 98% faster (1hr → 1min), deployment time 67% faster
See Details
  • Overview: Migrated from single VM to Kubernetes and built Grafana LGTM Stack monitoring system
  • Duration: 2025.04 ~ 2025.09 | Infrastructure owner
  • [Problem]
    • Single instance + Shell Script deployment → SPOF, manual intervention for recovery
    • No monitoring, relied on customer support for incident detection (avg 1 hour) → risk of users leaving the service
  • [Key Contributions]
  • [Results]
    • Eliminated CS dependency, incident detection: 1hr → 1min (98% faster)
    • Rollback time: 12min → 2min (83% faster)
    • Deployment time: 15min → 5min (67% faster)
    • Zero-downtime deployment and auto-scaling achieved
  • Tech: Kubernetes, GKE, ArgoCD, Jenkins, Grafana, Loki, Tempo, Prometheus
See Details
  • Overview: Added location-based property recommendation feature to existing real estate AI service ‘Butogi’
  • Duration: 2025.08 ~ 2025.11 | 3-member team, backend owner
  • [Problem]
    • Existing AI service only had real estate Q&A RAG, no property recommendation feature
    • Unable to handle location-based natural language queries like “Recommend apartments near Yeouido Station”
  • [Key Contributions]
    • Automated POI (Point of Interest) indexing to Elasticsearch via Spring Batch daily job
    • Built RAG pipeline: LLM extracts location & condition keywords from natural language → Elasticsearch similarity search to identify location → radius & condition filtering → property recommendation
    • Built vectorization and similarity search for property data using OpenAI Embedding API
  • [Results]
    • Launched AI property recommendation with natural language location search
    • Supported complex queries like “2-room within 10 min walk from Gangnam Station”
  • Tech: Kotlin, Spring Batch, Elasticsearch, OpenAI API, RAG, Geo-search, Embedding

Hello Fintech Co., Ltd.

P2P Online Investment Platform (₩2.5 trillion cumulative loans)
2023.03 ~ 2025.03 (2 years) | Backend Developer (5~10-member dev team)

Tech Stack: Java Spring Boot Spring Batch JPA Redis MySQL Kubernetes

  • Back-office Server Performance Optimization & Troubleshooting

    SQL tuning and caching, API response 87% faster (10.3s → 1.3s)
See Details
  • Overview: Migrated PHP legacy back-office to Spring with performance improvements and incident response
  • Duration: 2024.04 ~ 2025.03 | 2~6-member team, performance optimization & troubleshooting
  • [Problem]
    • JPA + MyBatis mix + OSIV enabled → 2 connections held per request → HikariCP Deadlock
    • Replication lag when reading from Slave immediately after Master write → data inconsistency → increased user inquiries
  • [Key Contributions]
  • [Results]
    • API response: 10.3s → 1.3s (87% faster), cached API: 5.1s → 1.3s (75% faster)
    • HikariCP Deadlock and Replication lag fully resolved
  • Tech: Spring Boot, Java 11, JPA, Redis, MySQL

  • Infrastructure Operations & Maintenance

    Covered DevOps gap, resolved session loss with Redis Session Clustering
See Details
  • Overview: Cloud-based server infrastructure operations and maintenance
  • Duration: 2023.12 ~ 2024.09 | Solo infrastructure owner
  • [Problem]
    • Infrastructure operations gap due to DevOps departure
    • K8s Ingress Sticky Session causing session loss on app restart
  • [Key Contributions]
    • Owned infrastructure operations until new DevOps hire
    • Managed 19 instances, 10 nodes, 3 web servers, 9 WAS, 2 DBs
    • Implemented Redis Session Clustering to resolve session loss
  • [Results]
    • Stable infrastructure maintained during gap period
    • Session loss resolved, improved user experience
  • Tech: Kubernetes, Jenkins, ArgoCD, ELK, Prometheus, Redis, MySQL

  • Spring Batch Performance Optimization

    Chunk/Partitioning, batch processing 64% faster (13min → 5min)
See Details
  • Overview: Migrated PHP/Crontab batch to Spring Batch with optimization
  • Duration: 2024.05 ~ 2025.03 | 4~6-member team, batch system migration & optimization design
  • [Problem]
    • Metadata Table Deadlock on concurrent Job execution – SELECT shared lock → INSERT exclusive lock transition caused deadlock
    • Tasklet with entire job in single transaction → MVCC snapshot missed intermediate deposit data
  • [Key Contributions]
    • Resolved Deadlock – compared 4 approaches (prevent concurrent execution, change isolation, DAO override, 5.0 upgrade), changed Isolation Level via Bean config
    • Chunk + 10 partition parallel processing – tested thread pool sizes, applied optimal value, documented and shared internally
  • [Results]
    • Batch processing: 13.3min → 4.8min (64% faster)
    • Transaction lock hold: 22min → 0.01s (consistency issues 99.9% reduced with Chunk commits)
  • Tech: Spring Batch, Java 11, JPA, QueryDSL

Side Project

Upvy - Educational Short-form Video Platform

2025.10 ~ Present | App Store | Play Store | GitHub

  • Overview: Educational short-form platform that turns scroll time into learning time
  • Tech: Kotlin Spring WebFlux R2DBC React Native Vertex AI
  • Highlights: Solo development · Live on App Store · 79% test coverage
  • Key Implementation: AI-powered YouTube video auto-editing pipeline (Vertex AI Gemini + Speech-to-Text)

Certificate

Activity