Built in the Open
NkosiFelix.com shows real engineering execution, AI platform work, and operator-grade delivery, out in the open.
This site is not a static portfolio. It keeps changing: shipped features, hardened interfaces, recruiter-facing pages, trust gates, health checks, mobile QA, and regression-tested public claims.
The goal is simple: make capability measurable.
Execution Ledger has the dated, cited record.
Python style
- Type hints throughout. Functions and variables are annotated, using
from __future__ import annotationsso it's evaluatable without runtime cost. A signature alone should say what a function accepts and returns. - Ruff for linting, mypy for type-checking. Both run in CI as a hard gate -- a change that fails either doesn't merge, including every module in this portfolio's own backend.
- Comments explain why, not what. Well-named functions already say what the code does; a comment earns its place only for a non-obvious constraint, tradeoff, or unusual choice.
- Tests for behavior, not coverage numbers. A test demonstrates a specific claim -- "this rejects an unsupported claim," "this degrades instead of crashing" -- not just executes a line.
Repository structure
Each project gets its own repository instead of living in a monorepo of unrelated work, so a reviewer can clone one thing and run its tests without pulling in context they don't need. The public reference projects follow that pattern: acceptance-harness, airgap-messaging-ingest, shoplabs-lakehouse-lab, and careerassistai-scoring-pipeline are each self-contained, with their own README, tests, and CI config.
AI coding tools
Four frontier AI tools are part of the day-to-day workflow, each used for a different job:
- Claude (Claude Code) -- the primary agentic coding tool used to execute large implementation passes under explicit architecture, evidence, safety, and review constraints.
- In-editor coding assistant -- faster in-context work: targeted edits, refactors, and code review without switching to a terminal-driven session.
- Grok -- fast iteration and second opinions: sanity-checking an approach, exploring alternatives, or a quick answer without a full agentic session.
- Codex -- focused code generation and review, particularly for narrower, single-file changes.
The point isn't "which tool is best" -- it's the right tool for the task, with the result always verified by tests and type/lint checks, never trusted blindly.
Same discipline, AI-assisted or not
The same rules apply whether a line came from a human or a model: a type signature, passing lint/type checks, and a test that demonstrates the claimed behavior. See Hiring Manager Proof for how that discipline shows up in this site's own architecture.