Selected work
Software, from the inside out.
I work across backend systems and the products they support. These are a few of the projects I’ve contributed to through El Cubo.
Product engineering · Data · Applied AI
Pinned
My work on Pinned connects messy information about places to a usable catalogue and product experience.
- Data intake and enrichment pipelines with field-level provenance and human review.
- AI-assisted categorization, spending controls, and evidence-based address resolution.
- Geospatial matching and search-index synchronization.
- Venue onboarding, account ownership, and company billing.
- Application delivery, queue workers, and the infrastructure supporting those workflows.
The decisions that interest me most are where these pieces meet: what should happen when sources disagree, a job retries, or an owner changes a record while automation is running?
Infrastructure behind the product
I designed and built Pinned’s infrastructure around K3s on AWS, with workloads spread across three availability zones. OpenTofu defines the infrastructure, and Fleet manages deployments through GitOps.
The data layer brings together PostgreSQL and PostGIS, connection pooling with PgBouncer, Redis for background work, and a three-node Typesense cluster for search. My work includes storage, backup and recovery procedures, workload identity, and secret management.
I also built the delivery path: immutable application images, a migration job that gates new releases, and separate web, scheduler, and queue-worker processes. Prometheus, Grafana, and ClickHouse-based observability help connect application behavior to the systems underneath it.
Cloud platforms · Developer experience
Galaxy
I’ve worked as principal architect and a primary implementer on Galaxy’s platform modernization, alongside the engineering and infrastructure team.
The work spans a TypeScript control plane, application and database workflows, deployment orchestration, authorization, and the customer interface.
It also reaches into the legacy Go and Meteor systems. A migration has to account for the commands, requests, domains, and behavior that customers already depend on.
From the legacy runtime to a regional platform
I designed Galaxy’s cloud architecture and led its implementation with the infrastructure team: moving application workloads from an AWS ECS runtime toward Rancher-managed RKE2 Kubernetes clusters on OVH bare metal across the US, Europe, and Asia-Pacific.
My implementation work connects that infrastructure to the product: independent build and deployment workers, blue-green releases, readiness checks, traffic switching, and rollback with captured configuration. The migration layer preserves existing deployment requests and routes applications between the legacy and replacement systems.
The platform also provisions managed databases, coordinates backups, controls workload placement and scaling, and exposes metrics and ClickHouse-backed application logs to customers. Cost efficiency and reducing dependence on a single cloud provider were part of the design.
Open source · Technical leadership
Meteor.js
I helped lead the work toward Meteor 3.0: moving the framework away from Fibers and toward native async/await, while giving the community a way to migrate gradually.
That work involved technical direction, release sequencing, code review, and communication with downstream teams. It was a shared effort across the core team and community.
Working together
I work with teams through El Cubo.
If you’re building a product or working through a difficult system change, get in touch on LinkedIn.