Dear Career 2

Curated jobs platform with admin ATS workflow and ingestion pipeline

Next.jsDjangoPostgreSQLOCRDocker

Full-stack engineering across frontend product surfaces, backend API workflows, and ingestion architecture.

Feb 2026

Overview

A hiring operations platform for curated Thailand job listings, combining a public jobs site with an admin ATS workflow for intake, review, and publishing.

Product Context

Designed for teams that curate and publish job opportunities quickly while keeping quality and consistency across data sources and channels.

Primary Goals

Support reliable job intake from URLs, image uploads, and managed source fetches.
Provide admin controls for review and publishing to website/Facebook channels.
Keep public browsing/search fast and simple for job seekers.

Architecture

Next.js 16 + React 19 frontend for public pages and admin dashboard workflows.
Django API backend with jobs domain services, ingestion commands, and publishing endpoints.
PostgreSQL core data store with Docker Compose full-stack local environment.
RapidOCR + ONNX Runtime + OpenCV pipeline for image-to-text job intake.
Chromium + Node browser-fetch fallback for client-rendered source pages.

Responsibilities

Implemented admin workflow for create/edit/delete, source management, and approval flow.
Integrated ingestion paths for pasted URLs, OCR uploads, and scheduled/manual fetch commands.
Connected frontend admin API layer with backend auth headers and shared-secret configuration.

Challenge Deep Dive

Ingesting jobs from heterogeneous sources

Job data comes from static and client-rendered pages with inconsistent structure and metadata quality.

Added source registry + fetch strategy controls, including Chromium-based fallback for dynamic pages.

Improved ingestion success rate and reduced manual re-entry for complex source pages.

Speed vs quality in admin intake

Ops users needed fast content entry but still required clean, reviewable job content before publish.

Introduced OCR-assisted draft generation and kept approval/publishing as explicit final workflow steps.

Faster intake throughput with better consistency in published listings.

Quick Start

Docker full stack

docker compose up --build

Seed fetch sources

python manage.py seed_fetch_sources

Run ingestion

python manage.py run_fetches

Open admin dashboard

/admin

Outcomes

Operational ATS/admin foundation delivered with multi-path intake.
Public jobs discovery and admin workflows aligned under one product model.
Clear technical base for expanding sourcing and automation modules.

Learnings

Ingestion systems need flexible fetch strategies, not one parser path.
Editorial approval boundaries keep platform quality stable as intake scales.
Tooling around admin speed has to preserve auditability and review confidence.

Overview

A hiring operations platform for curated Thailand job listings, combining a public jobs site with an admin ATS workflow for intake, review, and publishing.

Product Context

Designed for teams that curate and publish job opportunities quickly while keeping quality and consistency across data sources and channels.

Primary Goals

Support reliable job intake from URLs, image uploads, and managed source fetches.

Provide admin controls for review and publishing to website/Facebook channels.

Keep public browsing/search fast and simple for job seekers.

Architecture

Next.js 16 + React 19 frontend for public pages and admin dashboard workflows.

Django API backend with jobs domain services, ingestion commands, and publishing endpoints.

PostgreSQL core data store with Docker Compose full-stack local environment.

RapidOCR + ONNX Runtime + OpenCV pipeline for image-to-text job intake.

Chromium + Node browser-fetch fallback for client-rendered source pages.

Challenge Deep Dive

Ingesting jobs from heterogeneous sources

Job data comes from static and client-rendered pages with inconsistent structure and metadata quality.

Added source registry + fetch strategy controls, including Chromium-based fallback for dynamic pages.

Improved ingestion success rate and reduced manual re-entry for complex source pages.

Speed vs quality in admin intake

Ops users needed fast content entry but still required clean, reviewable job content before publish.

Introduced OCR-assisted draft generation and kept approval/publishing as explicit final workflow steps.

Faster intake throughput with better consistency in published listings.