The constraint
Gate hardware sits on unreliable connectivity and guards are not technical. The system had to keep working with the internet down and stay usable on a single touchscreen.
04 — Projects
The build, the constraint, the outcome.
Every case study below states what the system had to survive — bad connectivity, messy data, non-technical users — because that is what actually shapes an architecture.
- Sectors
- Residential security, enterprise ops, analytics
- Typical timeline
- 6–16 weeks
- Handover
- Source, pipelines, documentation
Case 01 — Security & AI
Smart Security Hub
AI-based society security: vehicle detection, face recognition and a tamper-evident entry–exit log, running on edge hardware at the gate.
What we built
On-device inference for plate and face recognition, a local queue that syncs logs when the link returns, and a guard console reduced to three actions.
- Python · OpenCV
- Local queue + sync
- Role-based console
Result
Replace this block with the real figure once measured — entries logged per day, manual register time removed, or incident lookup time.
—Add your metric
Case 02 — Web & Data
Enterprise Web Platform
A business management platform: role-scoped dashboards, operational reporting and exports, built to stay fast on ordinary office machines.
The constraint
Large tables, many concurrent users, and staff on older browsers. Client-side rendering alone would have made every report load feel broken.
What we built
Next.js with server-side rendering and streamed tables, PostgreSQL with indexed reporting views, and permissions enforced at the query layer rather than the UI.
- Next.js · PostgreSQL
- Server-rendered reports
- Query-level RBAC
Result
Add the measured improvement here — report load time before and after, or hours of manual consolidation removed each month.
—Add your metric
Case 03 — Mobile
Mobile Application
A cross-platform app with authentication, offline-first cloud sync and a component-driven interface shipped to both stores from one codebase.
The constraint
Users work in places with patchy mobile data and expect the app to open instantly and never lose an entry.
What we built
A local-first data layer with a sync engine and conflict resolution, biometric login, push notifications, and automated store builds through CI.
- Cross-platform codebase
- Offline-first sync
- Automated releases
Result
Add adoption or reliability numbers here — installs, crash-free sessions, or sync failures eliminated.
—Add your metric
Case 04 — AI & Automation
AI Analytics Tool
An analytics layer that ingests inconsistent operational exports and returns trends, outliers and plain-language summaries for non-technical managers.
The constraint
Source files arrive in different shapes every month, and the people who need the answers do not write queries.
What we built
A Python ingestion pipeline that normalises and validates each file, statistical outlier detection, and LLM-generated summaries grounded strictly in the computed numbers.
- Python pipeline
- Outlier detection
- Grounded LLM summaries
Result
Add the time saved per reporting cycle, or the number of reports now generated without analyst involvement.
—Add your metric
Next step
Your project could be case 05.
Bring the constraint you think is unsolvable. Those are the briefs we want.