2026 Operations Case Study

Public Customs AI Operations Platform

Built the operating path from large-scale data loading to admin backend, production migration, and performance validation. The focus was not a single feature, but a service structure that can keep running.

~300M rows2022-2026 Korea Customs Service / MFDS data loaded and processed
5 domainsData, admin, operations, infrastructure, and documentation handled together
Auto refreshData consistency path between development and production DBs
Load validationProduction service load-test system designed and implemented

Connected Operations Flow

Designed the data, application, and infrastructure boundaries so source data could reach service/admin features and return as operations feedback.

01 · SOURCEPublic Customs DataImports, exports, HS codes, tariff rates, item classification, and MFDS data
02 · PIPELINEETL Load & RefinementStandardization, transformation, repeatable load procedures, and result checks
03 · DATABASEMariaDB Operations LayerStandard design, production migration, and dev/prod auto-refresh
04 · SERVICEAdmin & User FeaturesHistory, Q&A, notices, improvements, comments, and admin APIs
05 · OPERATIONSValidation & FeedbackLoad tests, container management, standard guides, and manuals

Core Work

DATA / DB

Large-Scale Loading & Standards

  • Loaded and processed roughly 300M rows from 2022-2026 customs/MFDS datasets.
  • Refined HS codes, item classifications, tariff rates, overseas classifications, and MFDS data.
  • Managed system-wide DB standards and operating criteria.
ADMIN / BACKEND

Operations-Oriented Admin System

  • Designed the admin POC and API structure.
  • Built user, chat-history, and ETL management features.
  • Designed and implemented Q&A, notices, improvements, comments, and backend APIs.
OPERATIONS

Migration & Environment Refresh

  • Migrated the production DB and refreshed operating systems.
  • Built an automatic refresh program between development and production DBs.
  • Prepared Docker image/container cleanup and integrated management drafts.
INFRA / QUALITY

Production Transition & Quality Checks

  • Reworked on-prem network configuration and coordinated authentication APIs.
  • Handled certificate, DB, and production transition issues.
  • Designed, built, and tested a load-test system for the production service.
ARCHITECTURE

Connecting Developer, Operator, and User Views

  • Considered developer changeability and standardization.
  • Considered operator visibility, recovery, and management efficiency.
  • Improved architecture around service stability and feedback loops.
DOCUMENTATION

Transferable Operations Knowledge

  • Prepared data request/result documents and meeting records.
  • Drafted operating-system standard guides.
  • Wrote Q&A test cases and admin manuals.

Work Results

Repeatable Data Operations

Standardized loading and processing procedures so large public datasets could be reprocessed and expanded.

Admin Features as an Operating Layer

Extended a lookup-oriented admin page into a layer for history, content, and user feedback management.

Reduced Transition Risk

Reduced environment gaps and pre-release performance uncertainty with DB migration, auto-refresh, and load tests.

Integrated Management Basis

Organized scattered Docker images and containers and designed a unified management direction.

User Feedback Loop

Created a structure to collect and manage service improvement requests through Q&A, notices, improvements, and comments.

Sustainable Handover

Converted individual operating experience into reusable team knowledge through guides, test cases, and manuals.

Technology & Deliverables

Managed not only implementation technology, but also documents and validation tools used in real operations.

PythonMariaDBDockerLinuxETLAdmin APIDB SyncLoad TestTest CasesOperations Guide
View Full Career →