PulseOps

About Project

PulseOps helps on-call engineers monitor infrastructure health, configure alerting, and resolve incidents across a multi-cluster environment. The core challenge wasn't visual complexity - it was making health, alerting, logs, incidents, deployments, and on-call coverage behave like one operational record instead of a set of disconnected tools bolted together.

My Role & Activities

• Competitive Analysis & Benchmarking - benchmarked Datadog, Grafana, and PagerDuty to understand where their information architectures converge by convention versus by necessity, which informed treating dark mode as a functional choice for extended monitoring sessions and avoiding an interaction pattern already used elsewhere in this portfolio
• Entity Modeling & Information Architecture - the highest-risk decision was locking a Cluster → Service → Instance model before any screen work began, then mapping explicit relationships across the product (which alert triggered which incident, which deployment is a suspected cause, who's currently responding); that single decision is what lets the product read as one connected system rather than a collection of disconnected views
• Designing for Data Density & Progressive Disclosure - with 21 service instances across four environments, I used layered disclosure throughout: aggregated rows that expand into per-cluster detail, a persistent incident panel that stays visible even in a healthy state, and a log stream built around a live/paused distinction so engineers can freeze context without losing the feed
• State & Escalation Design - designed explicit coverage-gap visualization into the rotation calendar, a multi-step escalation path from chat to phone to SMS, and a rollback flow that requires confirmation before an irreversible action - each traceable to a real incident scenario the product needed to explain

Outcome

The result is a product where a coverage gap visibly explains an unanswered incident, where a deployment's config change can be the answer to a regression three screens away, and where an engineer can move from alert to root cause to resolution without losing the thread - a reusable pattern for cross-screen consistency that's easy to promise and difficult to actually design.