I take on the Salesforce work that's hard to get right. Async processing that can't drop a record. Integrations that recover when a callout fails. Business logic where a wrong number costs someone real money.
I'm a Senior Salesforce Developer, 4x Salesforce Certified including Platform Developer II, and I've spent most of the last four years inside complex, high-stakes orgs in lending and financial services.
All in one production Salesforce org for a US private lender, from first loan request through closing, sale and servicing.
| System | What it does | How it's built |
|---|---|---|
| AI document intake & auto-allocation | Borrowers and brokers upload a pile of files; each one lands on the right loan checklist item | Batch(5) → S3 → AI classify → Queueable → LWC review |
| Closing document engine, two vendors | Closing packages generated straight from loan, borrower and property data | OAuth client credentials, payload mapping, version diff before regeneration |
| Loan sale settlement & QuickBooks engine | Balanced debit and credit lines for every whole-loan sale | Per-investor day count (actual/360 vs 30/360), pool variance approvals |
| Loan pricing & term sheet platform | Bridge, ground-up and term pricers that turn a deal into a term sheet | LWC pricers + external sizing API, repricing flags on guideline changes |
| Layered entity ownership | Follows ownership through LLCs owned by other LLCs | Per-loan snapshots, 100% validated at every layer, look-through rollups, approval gate |
| Paystring engine & investor rollups | Month-by-month payment history and live investor totals | Legacy history for 4,000+ loans merged with monthly status, event-driven rollups + nightly job |
Also shipped: Do-Not-Use screening sync · draw & holdback QuickBooks lines · rate lock and cash-out letter PDFs · record-sharing models · production data fixes across tens of thousands of records
Clean-room versions of the patterns above, written from scratch with fictional data, so you can read the code and run the tests.
| Repo | What it shows |
|---|---|
| apex-async-document-pipeline | Callout-enabled Batch Apex, AI classification, Queueable allocation with retry and backoff, LWC review screen |
| apex-oauth-integration-kit | OAuth 2.0 client credentials with cached tokens, 401/429 handling, Queueable polling, versioned JSON diff |
| apex-ledger-engine | Balanced double-entry lines, actual/360 vs 30/360 day counts, idempotent regeneration, variance approvals |
| apex-layered-ownership | Look-through ownership across LLC layers, cycle detection, 100% validation, immutable approved snapshots |
$ sf project deploy validate --test-level RunSpecifiedTests
✓ 91/91 tests passed coverage per class ≥ 75%
$ sf project deploy quick --job-id 0Af…
✓ Deployed. Mirrored to UAT.
→ data fix: backup written, scope proven, rollback ready- Governor limits shape the design from day one
- Bulk-path tests with 200 records, not one
- Backups and a rollback script before any production data change
- Org state verified after every deploy, even a cancelled one
Got a hard Salesforce problem? Message me on LinkedIn