Arc 01 · Software Engineer
Engineering Baseline for AI Work
Turn a software project into a reproducible experiment, with explicit data, environment, cost, and verification boundaries.
Exit capability: Run an AI experiment another engineer can reproduce and audit.
- Mapped lessons
- 6
- Published now
- 6
- Full-arc estimate
- ≈8 hours
- Last edited
- 2026-08-11
Lesson sequence
Live units open into complete labs. Planned units stay visible to show the dependency path, but intentionally have no detail route.
- 01
Draw the AI System Boundary
LiveSeparate deterministic software, probabilistic model behavior, data, and human judgment before choosing tools.
Design reviewFoundation55 min estimateArtifact: AI system boundary map
- 02
Reproducible Experiments, Not Reproducible Notebooks
LiveMake code, data, seeds, dependencies, and hardware assumptions inspectable.
Build labFoundation80 min estimateArtifact: Experiment manifest
- 03
Data Contracts for Models
LiveSpecify schema, meaning, time, lineage, and allowed use before training or retrieval.
Systems labFoundation70 min estimateArtifact: Versioned model data contract
- 04
Define the Measurement Before the Model
LiveTranslate a product claim into offline, online, safety, latency, and cost measures.
Concept labFoundation65 min estimateArtifact: Measurement contract
- 05
A First AI Workload Cost Model
LiveEstimate tokens, accelerator time, storage, and network movement before committing architecture.
Build labFoundation75 min estimateArtifact: AI workload cost model
- 06
From Experiment to Service Boundary
LiveExtract a tested model contract from exploratory code without copying notebook state into production.
Build labFoundation75 min estimateArtifact: Inference contract