What I've built.

Specialized sub-agent teams triage, diagnose, and resolve datacenter operational tickets without human intervention. 25+ reusable skills built on sub-agent delegation and handoff patterns.

Impact ·~50% reduction in manual resolution effort for covered ticket categories.

Most of the detail here is intentionally generalized — APEX runs against AWS internal systems and the specifics aren't mine to publish. Happy to discuss architecture in interviews and 1:1 conversations.

Designed and built the watch-event processing pipeline on Kinesis, Flink, Lambda, and DynamoDB. Powered the experimentation infrastructure for ad-slot patterns, frequency, and house-ad placement.

Impact ·$14M+ in incremental ad revenue.

Integrated IMDb's 65M+ record catalog. NextUp service on Spring Boot, SQS, ECS, API Gateway with 99.999% uptime and sub-150ms P99. 22+ A/B experiments on top.

Impact ·$30M+ annualized streaming revenue.

Automated $300M+ in private pricing deals per year. Companion Document Generation Service and 6+ features for the internal contract-sharing platform that executes ~60% of AWS annual revenue.

Impact ·Deal turnaround from 2 months to 1 week.

Java and Node.js microservices, heartbeat monitoring backbone scaling to 20M sensor records, and a Three.js 3D visualization. Contributed to US-DHS SAFETY Act approval.

Impact ·3 seconds off first-responder radio-in time per incident.

Training on a focused corpus, writing the routing layer carefully enough to understand its failure modes. The intellectual thread connects to APEX: agent systems route between LLM personas at the orchestration layer; MoE routes between sub-networks at the parameter layer.

Manages long-lived Claude agent sessions, autonomous notifications, task tracking, agent teams, custom skills, and pomodoro cycles from the terminal. I use it every day.

Off-the-shelf brokerage dashboards don't show me what I want. Concentration analysis, intrinsic-value framing, exposure breakdowns, and a few lenses I keep reaching for.

Pulls my feeds, formats them clean, sends to Kindle each morning. Less screen time, more long-form reading.

Node.js Alexa skill with finer-grained controls than the default reading experience. Built at George Mason while volunteering with the accessibility program.

Lit-up wooden toy houses, a DIY humidifier, small sensor experiments. Projects that aren't supposed to scale.