Skip to content

Selected Work

Deep dives into real-world engineering challenges: the problem, the approach, and what happened.

SORT: REVERSE CHRONOLOGICAL · 09 ENTRIES
01
ARCHITECTURE

Pipeline Platform: A Config-Driven Document Processing System

Orchestrator, workers, dashboard, auth, and an MCP bridge built around a single DAG model

  • Micronaut
  • Postgres
  • RabbitMQ
  • Vue
534 → 1
PROCESSORS COLLAPSED INTO ONE DAG ENGINE
02
ARCHITECTURE

Pipeline Orchestrator: One DAG to Replace 534 Processors

A config-driven DAG engine for court-document processing with terminal evaluation and publish-before-commit safety

  • Micronaut
  • Postgres
  • RabbitMQ
  • DAG
534 → 1
CONCRETE PROCESSORS COLLAPSED INTO ONE DAG ENGINE
03
ARCHITECTURE

Pipeline Workers: Ten Generic Stages That Coexist With a Legacy Schema

Generic sub-processors over RabbitMQ, with read-only entity scan to share a database without owning it

  • Micronaut
  • RabbitMQ
  • Hibernate
  • Postgres
10
GENERIC PIPELINE STAGES REPLACING 534 COURT-SPECIFIC PROCESSORS
04
ARCHITECTURE

Pipeline Dashboard: A Thin SPA and the API-Perimeter Lesson

A Vue 3 operator surface that grew from read-only observability to config editing without ever owning a mutation

  • Vue
  • TypeScript
  • SPA
  • Kubernetes
0
DEPLOYS NEEDED TO ONBOARD A COURT
05
INFRASTRUCTURE

Pipeline Auth: A Standalone JWT Issuer With a Zero-Downtime Cutover

A Micronaut auth service that decoupled the new platform from the legacy stack, and the dual-controller alias pattern that shipped it without an outage

  • Micronaut
  • JWT
  • JWKS
  • Kubernetes
0
MINUTES OF OPERATOR DOWNTIME DURING CUTOVER
06
VISION

MCP Bridge: Shipping the Data Plane, Deferring the Framework

Building MCP-aware tooling against a framework that was still picking its annotation style

  • MCP
  • Micronaut
  • Java
  • AI Agents
10
PIPELINE OPERATIONS ALREADY MODELED AS TOOLS
07
ARCHITECTURE

EventHorizon: The Ultimate Data Detection Engine

A polymorphic ingestion, detection, and alerting platform born from solving the same problem across four companies

  • Flink
  • Kafka
  • Micronaut
  • ML
08
OPTIMIZATION

CI/CD Pipeline Optimization

Reducing build times from 47 minutes to under 8

  • CI/CD
  • DevOps
  • Performance
  • Gradle
47min → <8min
BUILD TIME REDUCTION
09
INFRASTRUCTURE

EKS Cluster Migrations

Zero-downtime Kubernetes upgrade under deadline pressure

  • Kubernetes
  • AWS
  • EKS
  • Terraform
0
MINUTES OF DOWNTIME