Nkosi Felix, Senior Data Platform Engineer
A public record of shipped platform work, QA hardening, AI guardrails, and recruiter-facing project pages.
This page summarizes real, dated commit history for nkosifelix.com -- not a description of intended work, but a record of what actually shipped. The source citation for every claim below is this repository's own commit history, 2026-07-08 through 2026-07-12 (154 commits). The full case study is at July 8-12 Build Sprint.
Tone note: this is a proof surface, not a pitch. Every card below names its evidence source and its current verification artifact, and labels anything missing as a TODO rather than asserting it.
FastAPI backend, evidence-gated chat adjudicator, and provider health checks shipped and hardened across the full build window, including a July 11-12 backend security pass.
Evidence source: the FastAPI backend and deployment config, live provider health checks, and the evidence ledger/adjudicator updates.
Verification artifact: live /health and /chat/status endpoints on production.
Evidence-bounded chat assistant with deterministic routing, employment-sequence answering, and a dedicated Safety checks surface.
Evidence source: fixing chat over-rejection, bounding chat skill routing, redesigning the chat launcher and adding Guardrail Validation, building employment-sequence answering, and closing chat gaps found by live regression sweeps.
Verification artifact: the live Safety checks section and chat widget on the homepage; backend regression suite (pytest).
Mobile navigation, hero video, and capture-slice regressions found and fixed across a dedicated mobile hardening pass.
Evidence source: decoding visitor region and fixing nav/chat mobile CSS, leading the mobile hero with video, repairing and tightening mobile hero capture, stabilizing capture-slice assertions, and polishing hero video and limit states.
Verification artifact: frontend/scripts/capture-mobile-critical-slices.mjs (21-slice mobile visual QA harness) run against production.
Machine-readable assets, sitemap/robots generation, and browser-preview polish (favicons, Open Graph) for public and AI crawlers.
Evidence source: adding machine-readable assets, SEO, check scripts, and the capabilities skeleton; adding the favicon and polishing browser previews; routing the new favicon/icon/manifest assets.
Verification artifact: frontend/scripts/check-seo-a11y.js static gate (run in CI and manually before every deploy).
A checksummed evidence manifest, claims register, and Career Evidence Platform Trust Gate that fail closed on unsupported claims.
Evidence source: adding the Career Evidence Platform Trust Gate, evidence ledger/adjudicator updates, rewriting the trust page copy, and closing perception-layer QA gaps.
Verification artifact: Trust page, claims-register.json, and backend/scripts/validate_manifest.py.
A 17-cluster semantic evidence mapper for job-fit scoring, real Credly/Microsoft Learn credential verification links, and a mobile tap-target hardening pass -- all found and fixed through audit-first QA, not assumption.
Evidence source: semantic evidence mapping for the job-fit evaluator, fixing two live routing bugs, adding verified credential links, adding Microsoft credential verification links and credential badge cards, and hardening mobile proof UX.
Verification artifact: the live Certifications section (dual PDF + Microsoft Learn / Credly verification links); backend/tests/test_resume_match.py and backend/tests/test_main.py credential/job-fit regression suites.
commit-ledger-from-7a9590a4.txt -- the citation source for every claim on this page and the linked case study.July 8-12 Build Sprint → -- the arc from buildout to hardening, cited against the same commit history.