Software Engineering - APP Developer - AI - startup help CLAUDE CODE (Mound)
‹image 1 of 1›
offered in person
offered virtually
QR Code Link to This Post
I am a local MN staff software engineer retiring from Big tech, looking to work with local startups or anyone looking to get something running with AI.
The pitch: Most teams don't have an AI problem. They have a verification problem.
You handed your team Claude Code. Output went up. So did the slop — code nobody really reviewed, tests that assert nothing, and a codebase drifting somewhere no one can fully explain anymore.
I spent three and a half years solving exactly that inside PayPal's Android org: a monorepo touched by 500+ engineers, where AI-generated code had to merge safely every single day. The fix wasn't better prompts. It was a development lifecycle where AI generates and deterministic systems verify — correctness lives in the gates, never in the model.
That's 17 years of engineering judgment turned into a workflow your team can actually run. I'm doing it right now with a startup team of non-engineers who are shipping real product.
What I set up:
→ Agent orchestration on a real SDLC — planning, grooming, implementation, review — not ad-hoc chat sessions
→ Deterministic gates: static analysis, quality gating, architectural fitness functions, contract validation, and mutation testing (which proves AI-written tests actually catch bugs instead of asserting nothing)
→ CI/CD, branching strategy, and a release process built for AI-speed throughput
→ Observability and crash reporting, so you know what shipped and what broke
→ Context systems and self-documenting workflows — your repo gets better at being worked on over time, instead of decaying
Three ways to work together:
1. Embedded — I join your team for a stretch and onboard the process alongside you
2. Flat rate — I ship you the agents plus documentation, and your team runs them
3. Built in your repo — I implement the agents in your codebase, you review every step
Best fit: startups standing up a new codebase who want it right from commit one, and teams who adopted AI tooling and watched quality quietly slide.