Software Engineer - Microsoft · Commerce
Atlanta, GA
DanielFullerton.
Software engineer at Microsoft on the Commerce team. I own the near-real-time reconciliation and billing infrastructure behind a mission-critical, SOX-regulated commerce platform that processes billions of billing events a month and has to close each month at 100% accuracy. Day to day that means Scala and Spark on Azure, reconciliation pipelines, livesite on-call, and the security and cost work that keeps a financial system trustworthy.
Before Commerce I spent two years on Microsoft’s data platform - GraphQL data APIs, data governance and lineage, Spark-based data-quality libraries, and a security migration that moved hundreds of services off application secrets to managed identity. Earlier I built retail operations apps at T-Mobile (Java/Spring, Angular) and led API architecture for talent-management platforms at Randstad (TypeScript, Node.js, Google Cloud). B.S. in Computer Science, University of North Georgia. I lean heavily on modern AI tooling to move fast without giving up rigor.
Work
Selected work
Figures generalized - no customer names or internal identifiers.
- 01
Near-real-time charge reconciliation
Owner · Commerce billing platform
I own the reconciliation layer for a commerce platform that processes billions of billing events a month across nine processing units and multiple Azure regions. I built automated dropped-event detection and replay, and hardened the alignment and comparison logic that keeps open and closed orders in agreement.
~10k dropped events/mo → 0100% reconciliation accuracybillions of events / month - 02
Legacy pipeline shutdown & streaming cutover
Multi-quarter program
Led the cutover of hundreds of thousands of billing groups off a legacy consumption pipeline onto a new streaming-sourced path - standing up processing and reconciliation infrastructure across regions and running auditable, financially-tiered batch cutovers with disciplined rollback when discrepancies appeared.
100s of thousands of billing groups100% accuracy at closezero SLA breaches - 03
Livesite, reliability & cost
On-call DRI · platform operations
As the on-call directly-responsible engineer, I'm the person the pager wakes. I run calm, telemetry-driven root-cause analysis on production incidents, hold the line through the highest-stakes week of every month at close, and mentor teammates through their first on-call rotations. Alongside that I cut recurring cloud spend, reduced CI build time, and shaved reconciliation runtime so month-close validation lands hours earlier for every partner team.
telemetry-driven RCA~78% faster CI builds~$35K/yr cloud cost removed - 04
SOX ownership, security & compliance
Owner · SOX / Security
I own four SOX controls end-to-end for the platform - compiling the annual audit-evidence packages, leading auditor and tester walkthroughs to minimal follow-ups, and authoring a reusable, AI-assisted evidence workflow that makes the process repeatable year over year. On the security side, I brought ~95 cloud resources into network-security-perimeter compliance across public and sovereign clouds and delivered managed-identity auth uplifts across every production region.
4 SOX controls owned95 resources → compliant10 production regions
Projects
Built end to end
Designed, built, deployed, and run by me.
- 01
Lean Site Services
Freelance · Design, build, hosting & launch

I rebuilt a construction-services company's website end to end and launched it on their domain. The old WordPress site on shared hosting became a static Next.js site on Azure Static Web Apps, with quote and contact forms delivered by a serverless function and a Pay Your Bill checkout through Authorize.Net that takes live card payments. I ran the DNS cutover through Cloudflare with a tested rollback path, and every push runs the unit tests before deploying to production, dev, or a per-PR preview.
live card payments$0/mo hosting141 unit + 16 end-to-end testsNext.js · TypeScript · Tailwind CSS · Azure Static Web Apps · Azure Functions · Cloudflare · Resend · Authorize.Net · Vitest · Playwright
Visit leansiteservices.com
Stack
What I
work with
When My AI Agent Said It Worked and It Didn't
An agent says something worked in the same confident tone whether or not it did. Here are four silent failures I've run into, and why I think checking agent work needs more than one layer.
Read →- Rebuilding My Claude Code Voice Mode to Be Hands-FreeOctober 26, 2026 · 5 minutes
- Thinking About AI Workflows Like Data PipelinesSeptember 28, 2026 · 5 minutes
- How I Start Each Day With Claude CodeSeptember 27, 2026 · 9 minutes
- Keeping My Claude and Codex Usage in the Menu BarSeptember 24, 2026 · 4 minutes