5 engagements, chosen carefully.

A 5-year-old Visualforce page handled 80% of daily interactions and had grown to 3,400 unmaintainable lines. The codebase had grown too tangled to safely touch — every change carried real risk.

I refused the rebuild. The data model was sound — only the UI was choking. Decomposed the page into discrete LWC components and migrated them section by section over 8 weeks, keeping the Visualforce shell live until each piece was proven in production.

6.2s → 0.9sFirst contentful paint dropped from 6.2 seconds to under 1 second. The codebase shipped with a component library the in-house dev can extend independently. Total cost: a quarter of the original quote.

Zurich's North American member portal had grown incrementally for years. 3 audience segments shared confusing flows, key tasks took 7+ clicks, and the component library was duplicated once per segment with no shared primitives.

I refused to ship segment-specific portals on top of duplicated code. Built a single reusable LWC library with audience-aware configuration, then collapsed the 3 flows into one streamlined journey.

7 → 2 clicksMembers complete common tasks in 2 clicks instead of 7. The library is now Zurich's foundation for the next portal — every component documented, tested, and segment-aware.

40+ independent reps carried 5th Axis alongside competing lines. None had Salesforce access — the licensing math doesn't work for users who'd log in twice a quarter. Quotas lived in a spreadsheet and commission disputes were quarterly.

I refused to scope licenses for users who'd never use them. Built a custom quota object, exposed dashboards through an Experience Cloud guest portal (no license cost), and wrote an Apex service that pulled rep performance from order data nightly. Reps got a Monday digest with a personal link — no login required.

$0 license costCommission disputes stopped overnight. "Where are we against quota?" went from a 3-day spreadsheet pull to a Monday glance. License savings covered the build cost in 7 months.

LandGeeks was managing properties, owners, and documents across disconnected tools. The team re-entered data constantly, and no single place showed a property's full picture — owner history, files, location, related contacts.

Built a console-style Lightning App: custom property and owner objects, a dynamic map component on the Google Maps API, a file management panel, and a filterable property list with multi-criteria search. Architecture stays component-first so each panel is independently extensible.

4 tools → 1Daily workflow collapsed from 4 tools into 1 console. Owners are linked to their records, files attach in context, and the map answers location questions that used to require a separate lookup.

Viritopia sells living walls — buyers spend months researching before they commit. Their team was creating Opportunities for every qualified inquiry, bloating the pipeline. A prospect can be fully qualified but still 2 quarters from a real decision.

Introduced a custom Pre-Opportunity object upstream of the standard pipeline. Qualified prospects enter research-phase tracking with their own stage logic, then graduate to Opportunities only when there's a genuine business decision in motion. Standard object stays untouched, so existing reports keep working.

Pipeline clarityPipeline stopped reflecting wishful thinking. Conversion from Pre-Opportunity became a meaningful signal, and the founder stopped tearing the forecast apart every quarter.