Inspiration

On campus, the standard rate to type and print a CV at a cyber cafe is incredibly cheap, often around ₦300. However, the result is usually an outdated, error-prone Word document that completely fails modern Applicant Tracking Systems (ATS). I realized that to build a valuable software product, I couldn't just sell a "typing service"; I needed to build a career accelerator. I wanted to create an AI agent that takes the friction out of resume building by meeting users exactly where they are, WhatsApp while also providing a premium, live-rendering web studio for precision editing.

What it does

It is a dual-channel AI career agent. The journey starts on WhatsApp, where users have a natural, casual conversation with the bot. The AI acts as an elite recruiter, asking targeted questions and instantly upgrading casual statements into professional, action-oriented corporate language.

Once the data is collected, the user can log into the web platform via a seamless passwordless WhatsApp OTP (Magic Link). The web interface features a split-screen "Artifact" architecture: a chat panel on one side to continue refining details, and a live-rendering, ATS-optimized document preview on the other.

How we built it

The system is architected to use a single, unified AI core that serves two completely different presentation layers:

Backend & Framework: Next.js handles the webhook-driven backend and API routes.

AI Orchestration: The Vercel AI SDK powers the conversational logic and dynamic tool calling to extract structured JSON data.

Database & Memory: Upstash Redis handles the session memory. This allows cross-platform continuity, a user can start a chat on WhatsApp and seamlessly pick it up on the Next.js web client without losing any context.

Communication & Auth: The Meta WhatsApp Cloud API powers the mobile interface. The WhatsApp bot is also repurposed to handle secure OTP token generation, completely bypassing expensive SMS services for web authentication.

Challenges we ran into

Building a bridge between a mobile messaging platform and a live-rendering web application presented significant architectural hurdles:

Structured AI Output: Moving from conversational text to a perfectly formatted, ATS-compliant JSON structure required precise prompt engineering and tool-calling validation.

Cross-Platform State Management: Ensuring the chat UI and the live-rendering PDF component stayed perfectly synced required meticulous React state handling.

Local Testing: Wrangling Meta's webhook requirements and establishing persistent ngrok static domains to keep the development environment stable was a major operational challenge.

What we learned

Building this project pushed the boundaries beyond standard CRUD applications into high-level system architecture. Orchestrating LLMs, managing streaming text, and dynamically handling tool calls provided a deep dive into modern AI integration. Furthermore, designing the WhatsApp-based OTP flow was a masterclass in thinking creatively about passwordless authentication, keeping the user ecosystem entirely enclosed, fast, and frictionless.

What's next

The immediate next step is expanding the web studio's rendering capabilities. The platform will soon introduce a highly aesthetic, visual layer inspired by the "Cozy Web", allowing users to generate tactile, scrapbook-style personal landing pages alongside their standard corporate CVs.

Built With

• nextjs
• ngrok
• typescript
• whatsapp-api