Lagos, Nigeria · nedum.xyz · [email protected]
I own quality and release readiness for production software: browser, API, integration, and data-boundary testing across complex SaaS, backed by automation that gates real releases rather than decorating a report. Seven years in software engineering, the last several focused on senior quality engineering, release ownership, and product security research.
Lately most of that work is testing AI output: LLM features fail confidently, plausibly and inconsistently, so I test them against golden sets, measure precision and recall, require every answer to cite its source, and block prompt or model changes that make any result worse.
Responsible-disclosure work has earned public credit for 2 CVEs and 3 GitHub Security Advisories.
QA harness for AI that extracts figures from financial statements, the kind of tooling used in acquisition due diligence and financial analysis.
- LLM evaluation: golden-set testing with precision, recall and grounded recall across six labelled balance sheets and income statements.
- Hallucination detection: every extracted value must quote its source line, and a citation that does not contain its value is caught as fabricated.
- Financial data validation: statements in thousands or millions, ambiguous dates, and balance sheets that do not foot.
- Regression gating: a release gate that fails any prompt or model change that makes a single field worse, run in CI on every commit.
- SQL data-quality checks over a diligence data room: duplicate uploads, missing checklist items, cross-source reconciliation, and brand-asset mismatches.
- API testing: FastAPI service with a Postman collection; 343 pytest tests.
- AI-enabled QA: LLM output evaluation, golden datasets, precision and recall, hallucination and fabricated-fact detection, regression testing of prompt changes.
- Release ownership: measurable acceptance criteria, explicit release gates, rollback conditions, and post-deployment verification.
- Test automation: Playwright, Cypress, pytest, and Python/TypeScript harnesses running deterministic integration, API, and failure-path checks in CI.
- Data and integration testing: SQL reconciliation, data-quality checks, and approval-workflow and permission boundaries across integrated systems.
- Product security research: authorization, OAuth, multi-tenant boundaries, integrations, billing state, and cross-layer business rules.
- Shipped With Bugs: independent QA and responsible-disclosure research on B2B SaaS.
Toolkit: Playwright · Cypress · Selenium · pytest · Postman · REST APIs · SQL / PostgreSQL · Python · JavaScript / TypeScript · Java · FastAPI · LLM APIs (OpenAI-compatible, DeepSeek, Claude) · Docker · GitHub Actions · Jira
| Portfolio | nedum.xyz |
| Research | shippedwithbugs.com |
| chukwunedum-davidson | |
| [email protected] |
About this repository: source for nedum.xyz
A static Parcel build deployed on Vercel. Requires Node 22 or
later (see .nvmrc).
npm install # install dependencies
npm start # dev server with hot reload, opens in browser
npm run build # clean + production build into dist/Layout
| Path | Purpose |
|---|---|
src/index.html |
Page content and structure |
src/css/ |
Sass sources |
src/js/ |
Page behaviour |
src/img/ |
Images and assets |
dist/ |
Build output (generated, not committed) |
Deployment. Vercel builds master and serves dist/. The custom domain is pinned
by CNAME, and vercel.json permanently redirects the nedu-m.vercel.app fallback
hostname to nedum.xyz so there is a single canonical origin.
Licence. The LICENSE file carries the MIT licence of the
Codrops template this site was originally built from.
Site content and copy are my own.