Tech Stack
The Platforms We Run On
Cloud-agnostic by design, we pick the right tool for each layer of the stack.
From bare metal to multi-cloud, we architect, migrate, and manage the infrastructure that keeps your business running at scale.
Infrastructure isn't glamorous — until it goes down. Then it's the only thing that matters. We build cloud and on-premise environments that you never have to think about: auto-scaling when traffic spikes, self-healing when nodes fail, and alerting before your customers notice anything.
Whether you're migrating a legacy monolith to Kubernetes, spinning up a multi-region deployment, or need a hybrid cloud that bridges your Hong Kong data centre with AWS and Azure, we design it, build it, and keep it running.
We cover the entire stack, from the physical layer to the orchestration layer.
EC2, Azure VMs, GCE: right-sized instances with auto-scaling groups and spot pricing optimisation
Docker containerisation, Kubernetes orchestration, Helm charts, service mesh with Istio
Cloudflare, CloudFront, Fastly: global content delivery with edge compute and WAF
VPN tunnels, VPC peering, private subnets, load balancers, DNS management
RDS, Aurora, Cloud SQL, Cosmos DB: automated backups, read replicas, failover
S3, Blob Storage, GCS: tiered storage policies, lifecycle management, cross-region replication
Zero-trust policies, MFA, role-based access, secrets management, compliance scanning
Prometheus, Grafana, CloudWatch, Datadog: alerting, SLIs/SLOs, incident response
GitHub Actions, GitLab CI, and Jenkins power automated build, test, deploy with rollback
Terraform, Pulumi, Ansible: infrastructure as code with drift detection and state management
Multi-region failover, RTO/RPO planning, and automated snapshots
Lambda, Cloud Functions, Azure Functions: event-driven compute with zero idle cost
Cloud-agnostic by design, we pick the right tool for each layer of the stack.
Every deployment follows a repeatable, gated pipeline: code to production in minutes, not days.
Code pushed to version control triggers the pipeline automatically
Docker image built, dependencies resolved, artifacts compiled
Unit, integration, and security scans run in parallel
Deployed to staging for smoke tests and approval gates
Blue-green or canary deployment to production with auto-rollback
A four-nines SLA target means roughly 53 minutes of unplanned downtime per year. Below is what one year looks like at that standard: the green ramp is healthy, a dim red cell is a brief degraded window, and solid red would be a full outage. Every cell is one day.
Cloud POS platform — Rust API, edge-delivered admin, and a safe gated deploy pipeline
A POS that takes real payments can't ship a bad build or a half-applied database migration to live tills. It needed deployment that's global and fast for the admin app, close to Hong Kong for the API, and safe enough that a broken release never reaches the shop floor.
The Rust/Axum API is containerised and deployed to Fly.io in the Hong Kong (hkg) region; the React admin ships to Cloudflare Pages for global edge delivery, with item and receipt assets on Cloudflare R2. A GitHub Actions pipeline runs the database migrations first and gates the rollout on their success, so Fly aborts the deploy if anything fails.
On-demand IT-engineer marketplace, built cloud-native on the edge
A real-time, two-sided marketplace (customers booking verified IT engineers with live GPS tracking, instant payments, OTP sign-in and push notifications) had to launch lean, scale on demand, and stay low-latency across Hong Kong, with no servers to babysit.
Built fully serverless and cloud-native: a Cloudflare Workers + Hono API running at the edge, Supabase (Postgres + realtime) for data, with BBMSL payments, Twilio OTP and Expo push. No VMs or containers to manage; it scales to zero between jobs and back up automatically under load.
EV charging network — serverless on the edge, with real-time telemetry from every charger
An EV charging operator needs a backend that's always on, cheap to run when idle, and able to hold live sessions to many chargers at once — streaming meter and status data in real time, without a fleet of servers to patch and scale.
Built serverless and cloud-native: a Cloudflare Workers backend at the edge, Neon serverless Postgres (with Drizzle ORM) and Supabase for data and realtime, and Cloudflare R2 for storage. Charger telemetry streams over WebSocket into the platform, and the driver app and operator console read it through one live source of truth — scaling with demand, no VMs to manage.
Let's audit your infrastructure and design a cloud architecture built for growth, resilience, and cost efficiency.
Book Infrastructure Audit