ChargeSmart Insight
Observability Dashboard
Built a real-time observability platform from scratch monitoring 70+ pipelines across 7 countries, integrated into the existing production stack.
70+
Pipelines monitored
Across 7 countries
Hours → mins
Incident detection
Via automated Slack alerts
2 weekends
Time to ship
End-to-end, from scratch
~90%
AI-written code
Using agentic coding workflows
About this project
Built a real-time observability platform entirely from scratch and integrated it into the existing ChargeSmart Insight production stack, covering data modelling, backend monitoring, a React dashboard, and Slack alerting.
The trigger was clear: incidents were discovered through manual checks and user reports, which meant failures sat undetected for hours. The goal was to make the system aware of itself.
I owned every layer: designed the data model to capture pipeline health across 7 countries, built the backend monitoring logic, wired up automated Slack notifications, and shipped a React dashboard surfacing live pipeline status and trends.
Shipped the entire system in 2 weekends. Around 90% of the codebase was written using AI agentic coding workflows, which compressed what would normally be a multi-week effort into two intense sessions.
The outcome was immediate: incident detection dropped from hours to minutes, data leaks and pipeline failures were surfaced proactively before reaching end users, and manual monitoring was eliminated entirely.
Tech stack
Key achievements
70+ pipelines monitored across 7 countries
Designed a data model and monitoring layer covering the full pipeline fleet, spanning every active market.
Hours → minutes incident detection
Automated Slack alerting replaced manual checks and user reports. The team now knows about failures before end users do.
Shipped in 2 weekends
Full system (data model, backend, dashboard, alerting) delivered in two focused weekends with ~90% of the code written via AI agentic coding workflows.
Proactive leak and failure detection
Surfaced data leaks and pipeline failures before they reached end users, turning reactive firefighting into proactive monitoring.