02 — Logistics · 2024
SkyTrack
Tracking experience and internal ops console for a logistics platform—live shipment state, exception handling, and tools dispatch teams used every day.
Role
Full-stack engineer
Timeline
12 weeks
Team
Product lead, 3 engineers, QA contractor
Status
Shipped
Support tickets (tracking)
−41%
Scan-to-UI latency
< 8s p95
Ops tasks / shift
+22% throughput
Uptime (launch quarter)
99.9%
The problem
Customers called support for status updates because the public tracker lagged behind warehouse scans. Ops relied on a spreadsheet to prioritize exceptions.
The solution
Introduced event-sourced shipment timelines, a real-time ops board, and a customer tracker that reflected the same source of truth within seconds.
What I owned
- Designed WebSocket fan-out and fallback polling strategy
- Built ops console views, filters, and bulk actions
- Integrated label providers and webhook normalizers
- Documented runbooks and on-call playbooks for ingest failures
Stack
- React
- Node.js
- Redis
- PostgreSQL
- GraphQL
- AWS
- Docker
- Datadog
Engineering highlights
- 01
Single timeline model
Every scan, delay, and handoff appended as events—UI and APIs read one stream instead of reconciling tables.
- 02
Ops-first UX
Keyboard-friendly queues, bulk reassignment, and saved views per hub cut average handle time.
- 03
Graceful degradation
When live channels dropped, clients fell back to polling with clear “delayed” states—no silent stale data.
How we shipped it
- 01
Audit & alignment
Shadowed dispatchers, catalogued failure modes, and aligned metrics with product and support leads.
- 02
Platform slice
Event schema, ingest workers, and read APIs. Load-tested fan-out paths before UI work.
- 03
Dual surfaces
Shipped ops console and public tracker in parallel, sharing components where it made sense.
- 04
Stabilize
Tuned alerts, back-pressure on workers, and training sessions for hub managers.