Why it was needed
01Political and bill data kept changing, and OCR research needed repeated evaluation, so collection and verification both had to be automated.
Civic data and OCR research tool
Politics King and OCR R&D
A project that combined civic-data product work with Korean OCR research, backed by APIs and internal tooling.
Civic DataOCRDashboard
PK
Civic Data / OCR
Data and research project combining civic-data automation with Korean OCR evaluation tooling.
Project overview
It mattered to build one research environment where data collection automation and OCR evaluation could keep moving together inside a small team.
Politics King and OCR R&D comes from a period where civic-data product work and Korean OCR research were being carried in parallel. The product side needed crawlers, APIs, and dashboard views, while the research side needed repeatable test-bed tooling and evaluation loops.
In the implemented scope, a Node.js API layer, Python-based crawlers, a statistics dashboard, Korean OCR benchmarking tools, and internal utilities were built together. It is an early project that shows service data work and research tooling in the same environment.
What I built first
02I built the Node.js API layer, crawlers, a stats dashboard, a Korean OCR testbed, and internal tools directly.
Where this approach works well
03This fits data products and research environments that need crawling, internal analytics, and repeatable evaluation tooling together.
Technical details
Tech stack
Data collection
Node.jsPythoncrawlersschedule-based ingestion
Service and visualization
API serverdashboard UIstats aggregationrecommendation logic
Research tooling
Korean OCR evaluationtestbed utilitiesbenchmark datasetsinternal admin tools
Services and tools
Crawler jobsAPI serverdashboard analyticsOCR testbed
Product surfaces
Selected screens
The screens below are the actual interfaces that defined the product experience.
Contact
If you are considering a similar product, feel free to reach out.
A short brief, a reference link, or the current problem definition is enough. I can help structure the right first version from there.