Abstract Security Pipeline Builder
Client project: Abstract Security (Crest Data Client Engagement)
Technical Deep Dive
Designed a React/TypeScript component system with drag-and-drop node graphs, nested conditional rule blocks that filter, aggregate, transform, and enrich (FATE) security data streams, an AI rule builder, and RBAC-gated controls, backed by 20+ telemetry integrations and 50+ typed FastAPI endpoints.
Client Context
Security teams route logs from many vendors (Datadog, ServiceNow, AWS, CrowdStrike) through a middle layer before they reach destinations like S3 or Splunk, where customers pay per volume ingested. Abstract Security needed a no-code builder so those teams could dedupe, filter, and enrich streams without writing streaming code, from zero to GA.
Execution
Built the drag-and-drop node graph, nested conditional rule editing stored as an ordered JSON structure the UI and backend both relied on, an AI rule builder, and a library of reusable rule blocks shareable across the org or kept personal under RBAC. Wired 50+ FastAPI endpoints behind typed Pydantic contracts with shared authn/authz, built 20+ telemetry source integrations, and designed a templating layer that turned each new connector from a custom build into a declarative config.
Outcome
Shipped zero to GA and was adopted by the full customer base at launch. Connector delivery dropped from 4 weeks to 1.5 weeks, runtime client errors fell 60% (50 → 20 per week), and React Query cache partitioning, cursor pagination, and debounced abortable fetches held sub-100 ms interaction latency under peak production load.
Why It Mattered
Every vendor used to be its own custom build. Turning that into a template plus config meant one engine could serve many consumers, and each edge case folded back into the template made the next integration easier instead of harder. It is also where the ownership grew, from assigned tickets to owning the builder's front-end architecture and the API contracts behind it.
“Architected and shipped a no-code security-pipeline builder from zero to GA, adopted by the full customer base at launch; a connector templating layer cut integration delivery from 4 weeks to 1.5 weeks and runtime client errors by 60% (50 → 20 per week).”
Core Stack
Metrics
integrations
20+
endpoints
50+
delivery_time
4 → 1.5 wks
error_drop
60%
latency
<100 ms