Things I've built — from tools for grassroots communities to systems that move data for thousands of users. Each one shipped, measured, and still running.
A multi-tenant point-of-sale system for cafés and small F&B businesses. The cashier keeps working when the internet drops and syncs once it returns, and owners see recipe cost and profit per product next to cash flow. It took real transactions on its first day in production (21 Aug 2026) and is now used by 4 F&B businesses.
A national teacher literacy and numeracy movement whose every content change depended on WordPress and a developer. I moved the site to Next.js and Payload CMS in 11 days, so staff now build pages themselves in an Indonesian-language dashboard, and added a Teacher Resource Library with 79 PDFs, 6 videos, and 20 interactive activities.
The official tourism information site for the Bogor City Tourism & Culture Office. The city's tourism information lived inside a chatbot iframe that Google could not index. In three phases over 14 days I took search-indexable pages from 0 to 76, built a back office where staff manage 55 destinations themselves, and added a directory of 250 official health facilities.
A community literacy movement in Labuan Bajo that existed only on Instagram. I built its first website in Indonesian and English, a 17-question needs-assessment survey designed for phones, and a back office where the team can read results without exporting anything.
Replaced a Wix site with one the foundation's non-technical board can update on its own, built in 4 weeks across two releases. It brings 15 years of program records into one structure, documenting 10 Gerakan UI Mengajar cohorts, and measures visits without cookies or third-party scripts.
Engineered a Python/Django Command Center dashboard for real-time health tracking and automated reporting.
Reduced manual reporting time by over 75%, establishing a centralized issue-logging database to accelerate bug resolution.
Boosted support workflow efficiency through automated PDF generation and helpdesk analytics.
Deployed a containerized environment replacing local static files.
Developed a full-stack web application enabling menu management and an expenses module.
Reduced order processing time from 5 mins to <1 min, and decreased order and expenses errors from over 10% to near-zero.
Integrated generative AI and NLP to automatically extract sentiment, keywords, and executive summaries from complaints.
Engineered robust two-way Google Sheets sync and Mekari Qontak API integration for seamless data pipelines.
Implemented AI clustering using Gemini and Groq to calculate impact metrics and auto-generate product recommendations.
Mapped an undocumented 5–7 year old PostgreSQL schema from scratch — no ERD, no documentation, just production data.
Filtered 162 total tables down to 77 actively used ones using pg_stat_user_tables, isolating real production entities from legacy noise.
Extracted all foreign key relationships via information_schema, producing a complete relational map of the system.
Generated both .mermaid and .dbml files — immediately usable in mermaid.live and dbdiagram.io.
Wrote a multi-CTE SQL query against an undocumented schema to extract and reconcile 2 years of payment transactions across 8 schools.
Covered Jan 2024 – Apr 2026 across 8 schools, handling partial payments, multi-billing transactions, and variable admin fees per payment method.
Cross-validated against the payment gateway reference report with a 99.76% transaction match rate — all 47 differences accounted for.
ERD mapping through SQL query to dashboard integration — done in under one working day, outside of job scope.
I'm currently crafting this section to be perfect.
Please stay tuned for updates!