Accra, Ghana. Building in public.
I am Ubong Udofia, Founder and CEO of Signcom Africa, an indoor digital out-of-home and retail media network in Accra, Ghana. Everything on this page is something I shipped, not something I plan to ship.
01 / About
I run Signcom Africa, an indoor digital out-of-home network in Accra. We put screens inside the places where people already stand still, pharmacies, salons, restaurants, gyms and retail floors, and we sell that attention to brands. That business is the reason I learned to build.
The screens are only half the job. The other half is everything around them: quoting, pricing with GRA levies applied correctly, writing proposals, producing the creative that plays on the screens, and following up on every lead. For a long time all of that sat with me.
Now it sits in software I built myself. I ship the whole chain end to end: the company website, a CRM with the rate card and tax logic built in, ad creative made with AI video tools, and an n8n automation that turns a raw enquiry into a priced, tax-accurate quote in under a minute. Nothing here is a mockup. Every item below is live, recorded or published.
Next I am taking the same pattern to other businesses in Ghana and Nigeria: find the repetitive work, build the thing that removes it, ship it.
02 / My AI stack
03 / Proof of work
Lets an advertiser see exactly where their ad will play in Accra: an interactive SVG map of the city, zone switchers, and the real venue and audience data behind each screen. Built as the company's own shop window.
Role: I designed, built and shipped it end to end, including the map and the venue data model.
Turns a phone enquiry into a priced, tax-correct proposal without a spreadsheet. Locally hosted CRM holding the rate card, GRA tax logic, proposal generation and an embedded AI assistant that drafts the client reply.
Role: Built the whole system myself, from the pricing rules to the AI assistant.
Runs locally on the business machine, so the walkthrough video is the proof rather than a public link.
A paper antenatal register records who came. It cannot record who did not. This shows a clinic exactly which expectant mothers have fallen off the WHO antenatal schedule, ranks them by follow-up priority, and puts a call or a message one tap away. The mother needs no app at all.
Role: Wrote the PRD, then designed and built the app from it.
The landing page is public. Sign in is invitation only, because it holds real clinic data.
The only mobile build
Tells a woman her exact foundation shade from one photo of her own face. Aurel reads undertone, depth and texture, places her on a sixteen step shade scale, and names the match: F12 Walnut, warm undertone, medium coverage, natural finish. It shows the step lighter and the step deeper beside it, scores how close the match is, and puts the products straight into a basket. About a minute, start to finish.
Platform: Mobile web app, runs on iOS and Android in the browser
A complexion brand sells sixteen foundation shades online. In a shop a customer holds the bottle against her jaw and knows in three seconds. Online she gets a grid of circles on a screen, so she guesses. The brand wants that counter moment back, on a phone, in under a minute, without a chatbot and without another quiz.
That is the brief. Aurel is my answer to it. I built it on my own initiative as a working product, and it can be fitted to any brand's own shade range.
Foundation is the hardest thing in beauty to sell online, because the one thing that settles the decision, holding the shade against your own skin, is the one thing a website cannot do. So the customer guesses. A wrong guess comes back as a return, returned foundation cannot be resold, and the support inbox fills with the same question every day: which shade am I?
Beauty brands selling complexion products direct to customers. Pharmacy and supermarket chains running a beauty aisle. Distributors carrying several shade ranges at once. Salons and makeup artists who want to size a client before anyone spends money. Anyone whose customer has to guess a shade from a screen.
Campaign work made with generative tools, produced end to end by one person. Each clip is embedded here and plays in place, muted, with controls.
Live campaign, published
Three Instagram Reels for a Ghanaian rice brand, each aimed at a different buying moment: the weekend family meal, the working woman who has stopped cooking, and the pack itself on the shelf. No crew, no studio, no shoot day. All three are live on the brand's account and play right here.
Role: Concept, script, presenter and scene generation, prompting, edit and delivery.
The weekend mealFor the family cooking their favourite rice dish at the weekend.
The professional who stopped cookingSpeaks to the corporate woman who has fallen out of the habit.
The pack on the shelfPuts the 1kg pack and the dishes it makes in front of the buyer.
Ad creatives and brand stills generated for screen placements and social. The three pension ads are one set built on a single insight, aimed at three different people: the employer who signs off the scheme, the employee whose salary is leaving every month, and the employer again on ground he already understands, health cover. The fourth is the still that sits behind the Masim campaign above, the pack and the dish in one frame. Click any image to see it full size.
Role: Concept, copy, art direction and generation.
Highest value build
An advertiser fills in a form. AI qualifies the lead and picks the tier, the score and the venue. A code node applies the Signcom rate card and the GRA levies. AI drafts the reply in the Signcom voice. The record is written to a Google Sheet, a Gmail draft is waiting for me to send, and the enquirer sees their recommendation on screen. Eight steps, end to end, live in n8n.
Turns a raw enquiry into a tax-accurate, ready-to-send quote in under a minute, with no manual pricing.
The rule that makes it safe to run: the model never touches the arithmetic. Claude and GPT qualify the lead and write the reply. The rate card lookup and the GRA levies, NHIL at 2.5%, GETFund at 2.5% and VAT at 15%, each on the net base, run in a code node. A quote can be wrong in tone and I can fix it. A quote that is wrong on tax costs money.
04 / My POD
We are a pod of builders who turned AI from a talking point into shipped work. In one week we went from ideas to live websites, working apps, and automations that run without us. We build for the businesses around us, in Africa and beyond, and we are only getting started.
The app the pod built together. Add what it does for the person using it.
My contribution: To be added
"5 Things Our Pod Does With Claude", the deck the pod presented at the end of the bootcamp. Drawn at random from real work: tax-ready quotations, a CRM with no developer, one channel and one voice, rebuilding a lost deck, and the time it said no when that mattered.
Anything else the pod shipped together.
05 / Certification
06 / Work with me
You get a working site or app on your own domain, live, not a design file.
The repetitive work runs itself, built with n8n and AI, so your team stops retyping things.
Campaign-ready spots and stills in days, without a crew, a studio or a shoot budget.
Your brand on screens in the places your customers already stand still, through the Signcom Africa network.