SabiCV

An AI career agent that rewrites Nigerian job seekers' CVs for the exact job they want — honest score first, ₦3,500, no fabricated claims.

Comment

"Sabi" is Nigerian Pidgin for know-how. SabiCV is an AI career agent that gives Nigerian job seekers the know-how their CV never shows: it reads their real CV against a real job advert, tells them honestly where they fall short — free — and then, for the price of a meal (₦3,500–₦10,000), rewrites their application the way a professional career coach would, without inventing a single claim.

The problem, and why this category

Nigeria's unemployment problem is not only a jobs problem — it is an access problem. Millions of qualified people lose openings they could have won because their CV doesn't speak the language of the advert, and professional CV services in Lagos charge ₦50,000+, out of reach for the people who need them most. Every interview a good candidate never gets is economic participation lost. SabiCV widens the doorway into employment: it prices professional application quality at street level and delivers it in minutes on a phone. That is what "fueling the tools that help economies thrive" looks like from the bottom up — not creating jobs directly, but unblocking the people who will fill and eventually create them.

What the AI actually operates

This is not a template shop with an API call inside. Gemini executes the core decisions of the business in production, and every decision is logged per order as evidence:

  1. Free gap analysis (the storefront). The agent scores the customer's actual CV against the actual job advert (0–100), names the top gaps and missing keywords. This honest, sometimes brutal score is our conversion hook — customers pay because the analysis already proved the agent read their CV.
  2. The rewrite. The agent restructures the CV toward the target role, translating real experience into the advert's language.
  3. The QA integrity gate — the agent auditing itself. A second pass traces every claim in the rewritten CV back to the customer's original. If it finds fabrication, it rejects its own draft and forces a corrective rewrite (up to three drafts, feedback accumulated). In our first end-to-end test the gate caught 24 fabricated claims and shipped a clean third draft. Career tools that hallucinate achievements get candidates blacklisted; ours refuses to.
  4. The truth receipt. Every delivered CV ship with a claim-by-claim provenance list — each statement traced to the line in the customer's own CV that supports it, marked "stated" or "inferred". The customer can defend every word in an interview.

The human founder does marketing, WhatsApp support, and payment confirmation. The agent does the product: analysis, judgment, writing, self-audit, delivery. That division has held for every order in production.

The business

  • Founded inside the window (first commit July 2026), solo founder, Port Harcourt, Nigeria.
  • Pricing: CV rewrite ₦3,500 · CV + cover letter ₦5,000 · Premium (adds LinkedIn rewrite) ₦10,000.
  • Payments: Selar (Nigerian merchant-of-record), so revenue evidence is third-party sales exports, not self-reported.
  • Revenue: ₦0 at submission — and we are reporting that honestly rather than dressing it up. The product, the agent pipeline, and the Selar checkout are live and verifiable today; four orders have been delivered end-to-end in production, including one where the QA gate rejected the agent's own draft and forced a corrective rewrite. Customer acquisition (WhatsApp, X, Nairaland) began on submission day. Expenses to date: ₦0 — Vercel hobby hosting, Gemini free tier, no domain purchased yet.
  • Unit economics: marginal cost per order is a few Gemini calls — effectively 100% gross margin at current volume; the constraint is distribution, not delivery capacity, which is exactly the constraint a solo founder with an AI agent workforce can attack.

Google Cloud

Gemini is the entire production workforce: analysis, rewriting, cover letters, LinkedIn rewrites, and the QA integrity gate are all Gemini calls with structured outputs. No other model is used.

Evidence in the submission

Per-order agent execution logs (every pipeline step, timestamped, including a production QA rejection and corrective retry), the live Selar storefront, and the live product itself — judges can run the free analysis on their own CV in under two minutes at sabicv.vercel.app.

Built With

Share this project:

Updates