Ari Hendrawan — Jakarta, ID

I build credit risk models, decision rules, and software for lending.

A Risk & Data Scientist based in Jakarta. My day-to-day work focuses on scorecard modeling and decision rules in Indonesia's multi-finance industry. Outside work, I build and ship web applications.

View projects My Resume
Portfolio

Selected projects

All 16 projects
Public-data credit-scoring benchmark

Independent public-data benchmark · UCI

ROC-AUC
0.7773
Ranking discrimination
PR-AUC
0.5518
Precision and recall
Brier score
0.1354
Probability error

31.5% of defaults captured in the top 10%.

6,000-row random holdout · Histogram gradient boosting.
Educational results; not employer-model or out-of-time performance.

Featured Risk & ML

Credit Scoring ML

Credit Scoring ML & Risk Modeling

My credit-scoring work covers preparing data, comparing models, and getting the inputs ready for handoff. Here, I use a separate public-data benchmark to share the results openly.

PythonLightGBMScikit-LearnRisk AnalyticsROC-AUC 0.777
Read the case study

An illustrative decision flow

  1. Input
    Check the request

    Validate the required fields before evaluating rules.

  2. Evaluate
    Apply the rules

    Run the versioned rules in their defined order.

  3. Explain
    Return the result

    Include the decision, its reason, and the rule version.

A system pattern, with no production rules or customer data.

Featured Risk & ML

CDE and GoRules Transformation

CDE & GoRules Decision Engine

I turn lending policies into GoRules decision tables, then check that the rules run in the right order and return the expected results through the API.

GoRules ZenFastAPIDecision TablesAudit EngineParity Tests
Read the case study
Document extraction and review
Fictional document
Name
Demo Person
Document
Sample only
Extracted fields
Name
Demo Person
Review
Required

KTP extraction + KK layout processing.
No real identity data. KK remains in development.

AI, OCR & Automation

Indonesian Document OCR

Indonesian Document OCR & Parsing

I built a service that reads KTP fields and flags anything uncertain for someone to review. I’m also working on a related pipeline for KK documents.

PyTorchLayoutLMField ExtractionHuman-in-the-Loop
Read the case study
A local demo of planning and photo memories. Activities are fictional; sample photos are AI-generated. Icons: OpenMoji, CC BY-SA 4.0.
Product Engineering

Koii

Koii: plans and memories for two

I built Koii as a private place for two people to keep a wishlist, plan activities, and save photos from time spent together.

Next.jsSupabaseGoogle Calendar APICloudflare WorkerWeb Push
Read about Koii
A local demo of the dashboard, transactions, and budgets. All amounts are fictional; no account or backend is connected.
Product Engineering

KasMate

KasMate: everyday money management

KasMate brings everyday finances into one place, with transactions, wallets, bills, and monthly budgets. I also built receipt entry and reports to help review the numbers.

ReactTypeScriptTailwind CSSPrivacy ControlsLocal-First
Read about KasMate

Technology Stack

The stack I use to build credit risk models, decision rules, and ship production systems.

PythonData & Systems
GoRulesDecision Engine
FastAPIREST APIs
Scikit-LearnCredit ML
DuckDBFast Analytics
PostgreSQLDatabase
DockerContainers
TypeScriptFrontend & Fullstack
SQL ServerPythonGoRules ZenLightGBMScikit-LearnDuckDBParquetFastAPIGitSQL ServerPythonGoRules ZenLightGBMScikit-LearnDuckDBParquetFastAPIGitSQL ServerPythonGoRules ZenLightGBMScikit-LearnDuckDBParquetFastAPIGit
TypeScriptNext.jsReactPostgreSQLSupabaseTailwind CSSDockerOpenCVLinuxTypeScriptNext.jsReactPostgreSQLSupabaseTailwind CSSDockerOpenCVLinuxTypeScriptNext.jsReactPostgreSQLSupabaseTailwind CSSDockerOpenCVLinux
Ari Hendrawan wearing a blue batik shirt
Background

A little about me
and my work.

I’m Ari Hendrawan, an Informatics graduate working in risk development in Indonesia’s multi-finance industry.

A lot of my work starts with questions about the data: how a feature was built, why model results differ, or why a decision came out a certain way. I work through those questions with business and IT teams.

Read more about my background
Experience

My work in risk development

Resume and certifications
September 2025 — PresentCURRENT ROLE

Risk Development Specialist

Indonesian Multi-Finance Company

I work on credit scoring, decision rules, and recurring portfolio analysis in Indonesia's multi-finance industry. I also work with business and IT teams to investigate data and application issues.

  • Credit scoring: Prepared historical credit data, checked candidate features, and evaluated how models performed on different samples.
  • Decision rules: Built and tested GoRules/Zen tables, including rule order, API response formats, and regression cases.
  • Portfolio analysis: Built data preparation and reporting workflows for portfolio risk and repeat-order analysis using Python, DuckDB, Parquet, and SQL Server.
GoRules ZenPythonLightGBMDuckDBParquetSQL ServerCredit Risk Analytics

GitHub Activity

80 contributions in the last year.

View GitHub
OctNovDecJanFebMarAprMayJunJulAugSepOct
Learn how GitHub counts contributions
LessMore
Contact

Get in touch.

Have a role or project in mind? I'd be happy to hear about it. You can also reach out if you're curious about something I've built.

Email: arihendra2802@gmail.com Connect on LinkedIn