Imanuel Puspa Wardaya: The Analytical Ledger

FRONT PAGE // LEAD STORY DISPATCH NO. 01-A VERIFIED FOR GENERAL RELEASE

Making data and systems genuinely useful.

Translating quantitative risk and portfolio dynamics into dependable scoring models, automated batch pipelines, and responsive local-first web applications engineered without network latency.

BY IMANUEL PUSPA WARDAYA / RISK & SYSTEMS DESK

FILED FROM JAKARTA, INDONESIA • OCTOBER 2026

Amid the persistent clamor for superficial automation and inflated AI promises, the foundational operational challenge of modern enterprises remains unchanged: ensuring data is parsed with integrity, financial risk models withstand rigorous validation, and deployed software reliably shoulders everyday human workloads.

My practice is grounded in a dual discipline: quantitative precision when dissecting institutional credit risk dynamics (from vintage delinquency maturation to score stability audits), paired with the craft of engineering local-first web systems that remain immediate, offline-resilient, and uncomplicated to operate.

Read Lead Feature ↓ Send Wire / Contact

FIG. 01: Imanuel Puspa Wardaya • Risk & Systems Editorial Desk, Jakarta.

OFFICIAL RELEASE • PWA VERIFIED ARTIFACT

SPECIFICATION: LOCAL-FIRST // INDEXEDDB + SUPABASE SYNC

FEATURED PRODUCTION SYSTEM

Takar

Know what you own. Buy when needed.

A progressive web application (PWA) for household inventory tracking and personal expense management. Engineered around a local-first architecture using IndexedDB for zero-latency interactions and offline resilience, backed by real-time cloud synchronization via Supabase and multi-workspace support (Personal & Household).

Item Inventory & Shopping Track household goods, storage locations, restock signals, and automated shopping list cost estimates.

Expense & Cash Sync Daily transaction logging, category summaries, and seamless one-click conversion from purchased goods into expense records.

Research Dossiers & Portfolio Audits

Methodological documentation on scoring models, vintage cohort analytics, policy thresholds, and automated ETL workflows.

FILE REF: MOD-02 // SCORECARD AUDITED

Credit Scoring & Segmentation

Developing and validating credit risk scoring methodologies, exploring regional model segmentation, and monitoring score distribution stability over time.

TECHNICAL NOTE [1]: Periodic score stability audits tracked via AUC and Population Stability Index (PSI).

Python
XGBoost
Optuna

FILE REF: MOD-03 // VINTAGE TIME-SERIES

FPD to NPL Vintage Curves

Analyzing the relationship between early payment difficulties (First Payment Default) and long-term non-performing financing via time-based feature validation.

TECHNICAL NOTE [2]: Early delinquency indicator modeling spanning 30-to-90 day cohort maturation.

Python
LightGBM
scikit-learn

FILE REF: MOD-04 // THRESHOLD POLICY

Risk Appetite & Cut-Off Bounds

Establishing an analytical framework to interpret risk indicators, contrast threshold boundaries, and document rational decision baselines for portfolio governance.

TECHNICAL NOTE [3]: Quantitative cut-off boundary design and expected loss provisioning scenarios.

Python
Pandas
statsmodels

FILE REF: MOD-05 // DATA PIPELINE ZERO-FAIL ETL

Automated Daily Report Pipeline

Designing recurrent data pipelines automating database extraction, figures consistency checks, executive document compilation, and timely delivery.

TECHNICAL NOTE [4]: Scheduled batch execution with automatic data reconciliation before distribution.

Airflow
Docker
n8n

FILE REF: MOD-06 // EXECUTIVE VIEW SELF-SERVE

Operational Metabase Analytics Suite

Crafting interactive dashboards to track delinquency trends. Designed, tuned, and validated high-performance analytical SQL queries and dashboard filters.

TECHNICAL NOTE [5]: Self-serve analytical dashboards powered by optimized Trino SQL warehouse queries.

Metabase
SQL
Trino

Note: Institutional risk modeling and analytics summaries remain high level to respect proprietary data ethics.

CLASSIFICATION OF ENGINEERING TOOLS & COMPUTING APPARATUS VERIFICATION REGISTRY: 2026

CATEGORY 01 // DATA SCIENCE

Credit risk modeling, machine learning, quantitative statistics, and hyperparameter tuning.

Python
SQL
scikit-learn
TensorFlow
XGBoost
LightGBM
CatBoost
Optuna
Pandas
NumPy
statsmodels

CATEGORY 02 // PIPELINE & INTEL

Recurrent workflow orchestration, containerization, distributed SQL queries, and dashboards.

Apache Airflow
Docker
Trino SQL
Metabase
n8n
FastAPI
Streamlit
GitHub

CATEGORY 03 // CLIENT & WEB SYSTEMS

Modern web applications, local-first browser databases, and cloud edge synchronization.

Modern JavaScript (ES Modules)
IndexedDB
Supabase
Progressive Web Apps (PWA)
Cloudflare Pages

Credit Risk & Data Analyst

PT Smart Multi Finance • Jakarta Headquarters

ACTIVE ROLE

TENURE: APRIL 2024 / PRESENT

PILLAR 01: PORTFOLIO & SCORECARD ANALYSIS

Evaluating creditworthiness distributions and borrower segmentations to mitigate potential default risks prior to contract underwriting.

PILLAR 02: EARLY DEFAULT & VINTAGE MODELING

Tracking the trajectory of first-payment defaults (FPD) through NPL across monthly vintage cohort curves.

PILLAR 03: RISK APPETITE & CUT-OFF DESIGN

Formulating portfolio risk tolerance boundaries and data-driven credit cut-offs that balance growth with capital reserve protection.

PILLAR 04: PIPELINE AUTOMATION & EXECUTIVE SUITES

Engineering high-performance Trino SQL analytical queries, self-serve Metabase dashboards, and automated Airflow ETL pipelines.

OFFICIAL CODE CONTRIBUTION LOG

Code Repositories & Commit Registry

26 contributions recorded in the last year.

Inspect GitHub (@notyourcircle) ↗

Learn how GitHub counts contributions ↗

GitHub contribution chart for notyourcircle

“Think clearly before modeling. Build with verifiable purpose.”

SPECIAL ESSAY BY IMANUEL PUSPA WARDAYA • FILED FROM JAKARTA

“A good system is not built to exhibit cosmetic complexity, but to transform the friction of real-world operational data into dependable, accountable decisions.”

To me, risk analytics and software engineering are not separated domains. They rest upon the same foundational discipline: rigorous honesty towards operational reality. The most intricate statistical model is useless if founded on flawed data assumptions, and a modern web application has no value if it stumbles the moment the network connection falters.

In my daily work across institutional credit risk, I regularly interrogate transaction-level anomalies, audit scorecard stability across economic cycles, and listen closely to operational branch teams so that risk governance policies can truly be executed.

When engineering client-side software like Takar, I bring that same ethos directly into the architecture: championing a local-first foundation with IndexedDB so users are never held hostage by network latency, layered with calm, dependable real-time cloud synchronization.

Respectfully filed,
IMANUEL PUSPA WARDAYA
Practitioner: Data, Risk & Web Systems

OPEN INQUIRIES

Let Us Exchange Ideas and Build Dependable Systems.

Whether you wish to discuss quantitative credit risk architectures, institutional ETL pipelines, or robust local-first web engineering, this telegraph wire remains open.

Send Wire Message / Email →

Official Dispatches: LinkedIn ↗ GitHub ↗ Instagram ↗