Problem

Admissions, timetables, exams and attendance were spread across spreadsheets and paper registers. Teachers were double-booked across classes, records went missing, and staff lost hours every week reconciling the same numbers by hand.

Solution

A full web platform covering student records, role-based staff accounts, timetable building, exam and question-bank management, and printable reports. The double-booking rule was pushed down into the database as a constraint, so a conflicting timetable entry is rejected at write time rather than caught later by a person.

Problem

The salon ran on a patchwork of disconnected tools and paper. Bookings came in by phone and message and were written into a diary, so double-bookings and no-shows were routine. Client history and preferences lived in staff memory. Nobody knew stock was low until a product ran out mid-appointment. Takings were totalled by hand, commissions were calculated at month end from scribbled notes, and the owner had no way to see which stylists or services actually made money.

Solution

One platform replacing the lot, with every module reading and writing the same database. Online booking with automated reminders closes the no-show and double-booking gap. A client CRM keeps visit history, preferences and formulas against each customer. Inventory deducts product as services are logged and flags low stock before it runs out. Billing and POS attribute every sale to the stylist who made it, which then drives attendance and commission automatically instead of month-end guesswork. A retention layer spots clients who have lapsed and triggers win-back campaigns. Because the modules share one source of truth, the owner gets a single dashboard showing revenue, staff performance and customer trends in real time.

Problem

Core business data was locked in a legacy Oracle 11g database while reporting and new development had moved to PostgreSQL. The nightly full export took hours and left analysts working on data that was up to a day stale. A straight lift-and-shift wasn't an option — the old system had to keep running throughout.

Solution

An incremental ETL pipeline orchestrated in Airflow, running every five minutes. Rather than reloading everything, it tracks a watermark on each table's updated_at column and moves only what changed — upserting where a primary key exists, appending where the legacy schema has none. Oracle 11g's older wire protocol needed the thick-mode client, so the whole stack was containerised with the Instant Client baked in, making the pipeline reproducible on any machine instead of depending on one engineer's setup.