[02]AI · Engineering productivity
AI-assisted software delivery
Making AI a delivery model, not a gadget: from a short brief to production code, with autonomous audits along the way.
- Organization
- YoLa Fresh
- Role
- Tech Manager — introduced and led
- Period
- Nov 2024 — Present
5–10×
faster production code delivery
15+
internal applications in production
30+
bots and AI agents in production
01The challenge
How do you ship much more production software, much faster, without giving up quality or control?
02My role
- 01Introduced AI-assisted development in the organization
- 02Tooling choices: Claude Code and the Anthropic SDK
- 03Set up autonomous code audits
- 04End-to-end delivery from a short brief
03The system
An augmented delivery pipeline
- 01Human
Idea
A business need
- 02Human
Specification
Intent, written down
- 03Agent
AI agents
Claude Code
- 04Agent
Code
Generated, reviewed
- 05Agent
Audit
Autonomous
- 06Agent
Tests
Before shipping
- 07Human
Deployment
To production
- 08Agent
Monitoring
Alerts < 60 s
Specification and sign-off stay human; AI agents write, audit and test the code.
04The execution
Technology
- Claude Code
- Anthropic SDK
- React
- Vite
- Vercel
- Node.js
- Python
- GitLab
People
People set the intent and make the calls; AI multiplies the execution.
Process
Specification first, then generation, autonomous audit, tests and release.
05The result
- Production code delivered 5 to 10 times faster
- Autonomous code audits built into the delivery cycle
- End-to-end delivery from a short brief