Hi, I'm Nick.
I simplify complexity and solve hard problems, fast.

Built and operated observability platforms across cloud-native services, Kubernetes, and AI/ML workloads. OpenTelemetry served as the translation and transport layer between workloads, collectors, backends, and alerting. I pushed teams toward native first-party instrumentation, and used the OTel Operator's auto-injection where it wasn't yet in place for faster visibility. Prometheus, Grafana, Alertmanager, Elasticsearch, and Jaeger, wired into Slack-based alerting and incident response.

Stood up incident command systems that moved response out of ad-hoc heroics and into a repeatable operating model teams trust under pressure. Defined roles, severity models, escalation paths, and Slack-based response channels; integrated alerting, observability, and timeline capture; and established postmortem practice and clean handoffs between engineering, platform, customer-facing, and leadership.

An LLM-assisted incident-response bot providing a conversational interface for alerting and incident management in Slack. It relayed operational alerts and answered questions during active response (what happened, when it started, what changed, what signals fired), keeping incident context in one place so the team can focus on diagnosis, mitigation, and recovery.

Production infrastructure for AI/ML-enabled services, the operational side of AI: making AI-backed systems deployable, observable, debuggable, and maintainable in real production. Cloud runtime support, deployment automation, and telemetry/logging/alerting wired into incident response, applying SRE and platform practices to practical MLOps on Vertex AI.

My small town had few attractions for families in the winter, so my family and I bought a beat-up bowling alley, and it became my real-world homelab for end-to-end technical impact. From unimproved land to operational in 90 days: an event-driven IoT + observability stack for legacy electromechanical equipment (ESP32 edge nodes, ESPNow, RS485 fallback, per-lane compute, MQTT), a React/TypeScript ops dashboard with live machine telemetry, and full ownership of architecture, reliability, cost, and P&L. I ran the technical side; the business closed in the summer.

A production AI voice agent for answering live customer calls for Lunar Lanes on Zoom Phone AI. It identifies itself as an AI on pickup, handles common questions in a consistent persona, and hands transactional requests (booking, lane availability) off to dedicated skills. I engineered the knowledge base from scratch (hours, two pricing modes with per-game vs hourly break-even math, the ownership/restoration story, payments, accessibility, conduct), and designed a stable persona so it carries calls that were never explicitly scripted, while keeping back-of-house policy out of public answers.

A job-search tracker that treats the hunt like a sales pipeline: applications are leads, sources and categories are segments, statuses are stages, and a live dashboard shows where effort actually converts. Every KPI and chart (conversion by source, pipeline funnel, response rate, salary ranges, in-demand skills) is derived live from the records, with applications going cold flagged automatically. No backend and no accounts: data lives entirely in the browser via localStorage, with JSON import/export and a one-tap PNG summary card. It ships as a single self-contained HTML file that runs fully offline.

The companion to Quarry. Where Quarry tracks your own pipeline, Assay analyzes the wider job market. A FastAPI and MongoDB backend serves a React/Recharts dashboard that breaks the market down by vertical (software engineering, data/ML, product), with endpoints for market pulse, job-board activity, in-demand skills, and salary bands. The mining theme is deliberate: a quarry extracts the ore, an assay tells you what it's worth.