Kwesi Owusu-Ansah

Software Developer

I build web apps, mobile apps and AI features like chat assistants and agents. At B Online I'm one of three developers, building products for several companies. On Luxity, South Africa's biggest luxury resale store, I built AI agents that sell to shoppers and bargain with people selling their items, and worked on the online store and the staff systems. I also build and run my own products.

Professional experience

B Online

Software Developer

January 2024 to present

I'm one of three developers at B Online. We build and run products for several companies, including Luxity, the Actuarial Society of South Africa and LLAAMI. Over 1,300 merged pull requests since January 2024. I scope features, build and ship them, and review my teammates' code.

Luxity

South Africa's biggest luxury resale store, online and in 4 shops

Over 900 of my merged pull requests are on Luxity.

  • AI sales agent. Built an AI agent that messages shoppers on WhatsApp a few minutes after they look at items on their wishlist and don't buy. The agent reads their past orders and wishlist, searches live stock, sends photos and a checkout link, and hands the chat to the right store's staff when a person is needed. Runs on head office's and all 4 stores' WhatsApp numbers, and as a chat on the website that adds items to the cart.
  • AI negotiation agent. Luxity buys its stock from people selling their items. Built an AI agent that bargains with a seller who turns down Luxity's offer, so staff don't have to. The agent stays inside price limits per brand and per item, never offers more than the head buyer has approved, and sends the buying team a summary when a deal is made.
  • AI product search. Built search that finds stock by meaning, not just exact words, updated every hour. The sales agent uses it to recommend items that are really in stock.
  • AI content. Built AI that writes each product's description, search title and social media preview text when the item goes live, and turns product photos into images of the item on a human model.
  • Automated messages. Built bulk WhatsApp campaigns sent from a spreadsheet, which the system checks before anything goes out. Built a morning message that sends each customer new stock matching their wishlist, starting with a short video of the best new pieces.
  • Store and staff systems. With two other developers, built the new online store that replaced WooCommerce: product pages, cart, checkout, search and filters. Built the online process for people selling their items to Luxity: the form, the agreement, the ID check and the payout. Built staff tools for sales, stock counts and transfers between stores.

Built with: Next.js, Laravel, Django, Claude, Gemini and OpenAI APIs, pgvector, AWS Lambda, Terraform

Actuarial Society of South Africa

Website assistant and staff tools for the professional body for actuaries

Main developer. Built the assistant and the admin dashboard, then reviewed the code of a second developer who joined later.

  • Website assistant. Answers members' questions from the society's website and documents, with links to the source. Points members to the right department when a question needs staff.
  • Always up to date. When a page on the website changes, the assistant's knowledge updates on its own. A nightly job removes anything the website no longer publishes.
  • Admin dashboard. Staff read chats, see what members ask most, manage documents and export chats each month.

Built with: React, FastAPI, PostgreSQL, pgvector, OpenAI API

LLAAMI

Product training for sales and frontline teams, in short lessons that feel like a game

Only developer. Built it alone, from the first version to today.

  • AI course builder. A manager uploads a product guide, SOP or menu. AI suggests the levels and quiz questions, and the manager changes them by chatting, for example “make it 5 levels”.
  • Learner app. Staff learn and take tests on their phones, with points, badges, daily streaks and a leaderboard.
  • Manager reports. Scores per person and team, filtered by date. Tests run weekly or monthly, and a missed test shows up in the report.

Built with: React, NestJS, Supabase, OpenAI API

Pace Online

Junior Developer

May 2022 to August 2023

Built mobile apps, websites and first versions of new products for clients.

  • SA Cargo Services. Built the first version of an Android app that port staff use to log every vehicle loaded onto and off ships, even without signal, with a web portal for the office. Port staff have now logged over 700,000 vehicles with it, across 3 ports.
  • National Institute for the Humanities and Social Sciences. Built the member portal and Android and iOS apps for a national community of doctoral scholars: profiles, groups, forums and events.
  • iSimangaliso Wetland Park. Rebuilt the website and built the mobile app for a World Heritage Site, with visitor information and online permit applications.
  • Msukaligwa Municipality. Built the first version of the residents' Android app: notices, tenders, service alerts, fault reports and bill information.

Built with: Flutter and Firebase for the apps

Own products

Wazi

AI customer support for online stores

  • Done for each client. I set up each store's agent, connect it to the store and keep it working.
  • Chat assistant. Answers visitors from the business's own website and documents, recommends products, and collects details from people who want a quote.
  • Business dashboard. The business reads conversations, sees the topics visitors ask about each week, gets a monthly report, and manages its team and billing.

Built with: Next.js, Hono, PostgreSQL, Cloudflare Workers, Stripe

Brevvi

Customer research reports from real online conversations

  • Research agents. Five AI agents research a market at the same time: pain points, desires, competitors and market data. They pull real quotes from Reddit, reviews, forums and YouTube, then group them into themes ranked by how often they come up, with source links.
  • Reports. Each report builds in the background and downloads as a PDF. Customers pay per report with credits.

Built with: Next.js, Hono, Inngest, Gemini API, Stripe

Technical skills

Languages
TypeScript, JavaScript, Python, PHP, Dart, SQL
Web and mobile
React, Next.js, Flutter, Laravel, Django, FastAPI, NestJS, Hono
AI
AI agents, RAG with pgvector, Claude, OpenAI and Gemini APIs
Data and infrastructure
PostgreSQL, Supabase, Firebase, Redis, Cloudflare Workers, AWS Lambda, Terraform, Docker, GitHub Actions

Education

Bachelor of Technology (BTech), Software Development

Nelson Mandela University

April 2021