← Back to projects
Fortune-telling service operator · 2024 · Commerce

Turning saju analysis
into something you can buy

An integrated management system for a saju service powered by generative AI. We built the operating flow end to end — saju algorithm development, precision analysis, HTML-to-PDF conversion and KakaoTalk alimtalk integration.

AI · Saju · Platform
4 months
To reach ₩1B in revenue
01 — CHALLENGE

Saju analysis as 'content' didn't grow revenue. Saju analysis as 'product' didn't have the operating infrastructure to scale.

Hand-read saju analyses are high-quality but expensive and limited in throughput. Automating opens scale, but analysis quality, result delivery and customer response all sat in human hands.

The operator didn't need 'AI to solve the analysis' alone — they needed the whole cycle automated, from analysis to delivery.

02 — APPROACH

We combined the saju algorithm with generative AI to automate the analysis itself, then wrapped result generation, delivery and notification into one system.

We built the algorithm alongside a saju expert, then layered generative AI on top — preserving precision while making expression more natural. Results turn into PDFs via HTML-to-PDF and dispatch through KakaoTalk alimtalk automatically.

AWS Lambda serverless architecture lets cost track traffic. The operator's admin shows orders, results and statistics on one screen.

03 — STACK

Stack.

01

React · Zustand

Order and result pages with lightweight state management.

02

Spring Boot · Node.js

AI processing and PDF conversion separated for operating stability.

03

AWS Lambda · RDS · S3

Serverless + RDS MySQL + S3 · CloudFront for static assets.

04

GitHub Actions · CodePipeline

CI/CD automation reduces deployment overhead.

05

ChatGPT

Generative AI layered over the saju algorithm to raise analysis quality.

Got a similar problem?
Let's build it.

Request a consultation →See other projects