Projects
What we have built for clients
Ten systems, built and running. Each one started from a real problem at a real business.
Shops and small businesses
Selling, stock, people, money owed — the work of the day.
Point of Sale & Store Manager
Built and running · not yet rolled out
- The problem
- A shop runs the till on paper or on a phone's notes app, so nobody can say at closing time what was sold, what is left, or whether the drawer balances.
- What we built
- A till screen with stock, purchasing, suppliers and reports behind it, cash accounted for by shift, and the trade-specific half — rooms, repair jobs, rentals, tables — switched on per shop.
HR, Attendance & Payroll
Built and running · not yet rolled out
- The problem
- Attendance is argued over every month: a punch nobody can verify becomes a payroll line nobody can defend, and the insurance and tax filings are rebuilt by hand from spreadsheets.
- What we built
- Six ways to clock in with the suspicious ones queued for review with their evidence, a timesheet that closes, payroll from a versioned formula, and the insurance and PIT filings off the same approved numbers.
Learning Centre
Shipped product · running
- The problem
- A training centre keeps classes, attendance, fees and coursework in four places that do not agree, and the staff web, the teacher app and the student app each need the same answers.
- What we built
- One API behind an admin web, two mobile apps and the LMS, so a change to a class or a payment shows up the same way wherever it is read.
Dealer Orders
Shipped product · multi-tenant SaaS
- The problem
- A manufacturer takes dealer orders over Zalo and tracks what each one owes in a spreadsheet, so an order can be missed and a debt can be argued about with no record to point at.
- What we built
- A multi-tenant ordering system: each dealer places orders against their own price list and sees their own balance, and the manufacturer sees all of it in one ledger.
Stock Count
Desktop app · install required
- The problem
- Stocktake is two people with a clipboard, and the count only disagrees with the books after everything has been put away.
- What we built
- A desktop app that counts by barcode and reconciles against the book figure as it goes, so the gap shows up at the shelf rather than at the desk.
Marketing and content
Getting the product in front of people, and keeping the content.
SEO Autopilot
In build · runs end to end on dev
- The problem
- A small team juggles six to ten SEO tools and still misses the obvious: queries sitting at rank 8–20, and pages nobody reads because nobody read Search Console this week.
- What we built
- An agent that runs the loop on a 7, 14 or 30-day cycle: read Search Console, score what is worth doing, plan, write, publish and measure — with a human on the approve button.
Livestream Control Desk
Shipped product · running
- The problem
- A livestream selling session needs one person on the encoder, one on the product list and one on the comments, and the three of them are in the same room shouting.
- What we built
- A control desk that drives OBS from another machine, pins products to whatever is on screen, auto-replies to comments, and can run a virtual host from a script.
Self-Hosted CMS
Self-hosted · admin and API built, frontend in progress
- The problem
- A content team pays per seat and per record for a hosted CMS, and the content they are paying to store is the one thing they cannot move out cleanly.
- What we built
- A self-hosted CMS built to take the place of Contentful and Sanity: the schema, the API and the data sit on infrastructure the client owns.
Prototypes
Built to test an idea. Not finished products.
Family Tree
Shipped product · self-hosted
- The problem
- Family-tree software is built around one marriage and two parents, so a real Vietnamese family — second wives, adoptions, a branch that moved south — does not fit in it.
- What we built
- A Vietnamese family tree that models the relationships families actually have, and renders a chart that stays readable as the branches multiply.
Parental Control
Interactive prototype · two paired apps
- The problem
- Parental controls are written as something done to a child: a lock the parent sets and the child works around, with no version of the rule the two of them agreed on.
- What we built
- Two paired apps — one on each phone — where screen time, web filtering and location are a rule both sides can see, and extra minutes are earned rather than begged for.
Have something like this?
Tell us the problem. If we can build it, you get the scope and the timeline before anything starts.
Book a consultation