Projects I've built.

What I built

A proprietary drag-and-drop web builder that lets the delivery team assemble client-ready landing pages and product mini-sites from a library of branded blocks. I designed the editor UX, built the component library, defined the JSON content schema, and wrote the live-preview engine.

My day-to-day work on this

• Block authoring: hero, columns, feature lists, CTAs, forms, footers — each as a reusable, themable HTML/CSS/JS module.
• Editor UX: drag-to-reorder, inline content editing, real-time preview pane, undo/redo, template save/load.
• Theming system: token-based design system (colors, type scale, spacing) so one engagement can re-skin without touching markup.
• Onboarding: wrote the internal docs and ran two enablement sessions so the wider team could ship without me.
• Product roadmap: prioritized features based on actual project requests; turned each "we need X for client Y" into a generic block.

What I learned

Treating internal tooling like a real product — with users, docs, and a roadmap — compounds team velocity in a way one-off scripts never will. The hardest part wasn't the editor logic; it was deciding what NOT to build so non-developers could still use it without breaking things.

What I built

Multi-page corporate brand and product sites on Adobe Experience Manager (AEM) — the enterprise CMS. Reusable templates and dialogs let the client's marketing team author new pages without engineering tickets. Multi-locale support means a single template ships in EN/ID/JP variants from day one.

My day-to-day work on this

• AEM Templates: built page templates and editable Touch UI dialogs the marketing team can fill in themselves.
• CMS Blocks Library: hero, feature lists, CTA bands, contact forms, footer variants — all configurable per brand.
• Brand-system implementation: turned brand guidelines (color, type, spacing) into design tokens consumed by every component.
• Multi-locale routing: language fallbacks, RTL support, locale-aware date/number formatting.
• Performance tuning: lazy loading, image sprites, critical CSS — Lighthouse 95+ on shipped pages.
• QA & cross-browser: tested across IE11 (legacy clients), Edge, Chrome, Firefox, Safari, mobile WebViews.

What I learned

Authoring UX matters as much as visitor UX. A perfectly designed page is worthless if the client's marketing team can't add a banner without calling support. Good dialog labels and sensible defaults reduce support tickets by an order of magnitude.

What I built

Customized Magento 2 storefronts for ERHA — Indonesia's largest dermatology clinic chain. The work covered seasonal campaign rollouts (banners, promo blocks, voucher checkouts), product catalog operations, and dozens of conversion-focused UI improvements over 14 months.

My day-to-day work on this

• Storefront themes: customized Magento front-end themes, layouts, CMS blocks, page templates.
• Campaign delivery: Ramadan, year-end, anniversary, flash sales — built dedicated landing pages, banners, voucher flows.
• Catalog operations: product uploads, pricing updates, content refreshes, SKU corrections — kept the catalog accurate across 1000s of products.
• Promo & voucher UI: coupon entry, applied-discount display, expiry warnings, T&Cs modals.
• Conversion improvements: simplified PDP layouts, sticky add-to-cart, trust badges, simplified checkout — measured win rates per change.
• On-site merchandising: reordered category pages around campaigns; A/B tested hero placements with the marketing team.

What I learned

In commerce, small UI moves change real revenue. A 3px tweak to a CTA button can show up in dashboards by the next morning. That feedback loop made me obsessive about measuring before/after on every change instead of trusting my eye.

What I built

A smart CV-screening platform that helps HRD teams shortlist candidates up to 80% faster. OCR extracts structured data (skills, experience, education, certifications) from any CV PDF — regardless of layout — then an AI scoring engine ranks each candidate 0–100 against the job description.

My day-to-day work on this

• OCR pipeline: reliable text + structured-field extraction from any CV PDF layout (multi-column, scanned, dense).
• AI scoring engine: ranks candidates 0–100 vs JD on skill match, years of experience, education, certification fit.
• HRD dashboard: upload batch CVs, filter by score, shortlist, batch-export, view per-candidate audit trail.
• Red-flag detection: auto-flags employment gaps, mismatched titles, missing required fields.
• Workflow integration: CSV/Excel export so shortlists drop straight into ATS or recruiter tools.
• Trust & transparency: every score shows the reasoning — which keywords matched, which gaps lowered the score.

What I learned

AI scoring only earns trust if the reasoning is transparent. HRD won't shortlist a 92% candidate they can't justify. Every score needs to show its work — which keywords matched, which gaps lowered the score, what the model actually saw in the CV.