Austin Lippert

Available

Senior Software Engineer · .NET & TypeScript

I take a vague problem and put working software in production, which I have done for twenty-plus years as a self-taught developer who was often the only engineer on the problem. Improving code is the part I enjoy most, and I measure what I claim.

Knoxville, TN|linkedin.com/in/austinlippert

Impact Snapshot

$15B+

Payment volume a year

  • Built the gateway from scratch; it became its own business, Payment Brands.
  • Per-brand tokenization kept each acquired brand out of PCI scope.
  • Joined Ministry Brands as employee #7 and watched it grow to 800 developers.

11–16 min→<20 s

Data pipeline cycle

  • Built the warehouse and pipeline from nothing, over 300M source records.
  • A container search that scanned the whole table now answers in 1.5 ms from a trigram index.
  • The delay turned out to be thirty seconds of metadata overhead, which came out in code.

7.91→9.53

Code health score

  • Split a 158 KB Python file into 60 modules, taking its type errors from 36 to 0.
  • The repo went from Warning to Excellent, and its worst file from 1.0 to 6.05.
  • Once the structure was in place, fixes per feature shipped fell from 3.3 to 0.5.

How I Deliver Now

I use AI agents every day, and while they make me faster, the judgment stays mine.

  • About a dozen agents run daily (Claude, Codex, Hermes, OpenClaw) across repo analysis, documentation and business work.
  • Each agent works inside the repo it is given, and nothing ships until the tests and gates agree.
  • Work that would be scoped in weeks lands in days, because the verification is automated.
  • I use whatever the team already runs: in one engagement I rebuilt a 38,000-line Blazor client as React/Next.js and rewrote its pipeline in Node, .NET stays my default, and I can learn a new project in hours rather than days.

Technical Skills

All of it shipped in production; whatever the team runs, I move it toward the current LTS.

Languages

C# · Python · TypeScript · JavaScript · SQL

Backend

.NET (any version) · ASP.NET Web API · Entity Framework · microservices · Node.js · Express · Windows services · Azure Functions · AWS Lambda

Front end

Blazor · Vue 3 · React · Next.js · Angular · Tailwind · component libraries

Data

PostgreSQL · SQL Server · Snowflake · MongoDB · Redis · ETL/ELT · CDC · index tuning · warehousing

Cloud & DevOps

Azure · AWS · Docker · CI/CD · Azure DevOps · GitHub · Jenkins

Quality & tooling

TDD · SOLID · DDD · code-health tooling · dependency and CVE remediation in CI · Vitest · Jest · Playwright · xUnit · NUnit · agentic development

What I Solve

Payments, compliance & fraud

Money moves at volume here, so every change is a risk decision.

$15B+/yr90+ brandsPCI scope cut per entityfraud risk 96.7% below average

  • Built the PCI-compliant payment gateway from scratch, processing $15B+ a year across 90+ brands. Per-brand tokenization kept each acquisition out of its parent's compliance scope. Ministry Brands, 2014–18
  • Added real-time fraud analytics: risk on monitored attack vectors sits 96.7% below average, and the framework cut fraud damages 60%. TransCard, 2011–12
  • The platform absorbed growth from 14 to 800 developers and 90+ acquisitions without adding payment risk.

Data platforms & pipeline performance

The numbers have to reach people fast enough that they act on them.

300M → 100M → 40M rows132 s → 1.79 s tick7 days → <5 min ETL

  • Built a logistics warehouse and pipeline from scratch: roughly 300M source rows, 100M ingested, 40M in gold, with log noise dropped at the ingestion boundary. Contract, 2026
  • Cut the pipeline cycle from 11–16 minutes to under 20 seconds, and tick latency from 132 seconds to 1.79 seconds, by removing 30 seconds of metadata overhead in code instead of buying a bigger server.
  • Cut a healthcare ETL from seven days to under five minutes, freeing six staff and creating five roles, and built the de-identified dashboards it feeds, where HL7 records are validated and anonymized upstream so patient data never reaches the analytics database, and what is shown is gated by role-based access.

Refactoring & code nobody dares touch

This is the code nobody wants to open, where every change is a gamble.

22 → 60 modulestype errors 36 → 0health 7.91 → 9.530 of 103 → 569 tests

  • Split a 158 KB single-file Python engine into focused modules: 22 to 60, type errors 36 to 0, 180 files in one pass, tests green the whole way. Contract, 2026
  • Code health rose from 7.91 to 9.53 (Warning to Excellent) across 1,700+ files, and the worst file climbed from 1.0 to 6.05.
  • The fix rate followed the structure: 3.3 fixes per feature shipped fell to 0.5, and the file we had patched 18 times in five weeks has needed two fixes since.
  • Took a client application from 0 of 103 tests passing to 569 across four pull requests, cutting runtime from 29 minutes to under 6 and surfaced seven production defects.

Cost, speed & delivery

The same work runs for less money, usually without buying anything new.

runtime −99.86%cloud cost −85%deploy −60%conversion +40%

  • Moved four legacy processes to cloud-native services on AWS: 99.86% less runtime for the same work, 85% lower cloud cost. InhabitIQ, 2021–22
  • Led the TFS-to-Git migration and the review process that came with it, making deployments 60% faster and taking deployment errors to zero.
  • Rebuilt a customer-facing application on a reusable component library, which lifted loan conversion by 40% and cut completion time by 81.25%; no dedicated front-end sprints since.