Engineer and architect with 11 years building cloud infrastructure, data platforms, and the backend systems teams rely on — including data pipelines that stream 35 million events an hour. I connect what the business needs to what the system should actually do.
Selected projects across cloud infrastructure, data & ML platforms, and backend systems — most still running in production. Filter by stack, then see independent work below.
A flexible distributed messaging system built on the CloudEvents spec and AWS, designed to support rapid iteration and evaluation of AI/ML models and backend services — with logging, tracing, and shadow testing built in.
Resilient AWS data pipelines — Kinesis Firehose and custom Lambdas — streaming security and telemetry from 17 accounts, 16 regions, and 4 clusters into a centralized observability platform with real-time alerting.
Terraform GitOps workflows and a Landing Zone with an account vending machine and security baselines — managing 30,000+ resources and migrating 270+ workspaces between IaC platforms.
Designed and own the technical controls behind the company's SOC 2 — access management, audit logging, monitoring, encryption, and change management across the cloud platform. Drove the Type 1 to completion and run the ongoing Type 2 audit and controls.
Architected an MLOps and data-science development platform on GCP, with per-user ephemeral Dataproc clusters and Terraform adoption across global teams. Presented at the 2019 Data Symposium.
Enterprise Jupyter and PySpark development platform on an on-prem Hadoop cluster, adopted by 30+ data scientists and enabling 20+ ML models in production.
Migrated an Elasticsearch cluster from a Cisco-managed instance to GKE with zero downtime, optimizing resource utilization and establishing monitoring.
A Python library ecosystem powering mission-critical FinTech products: secure REST clients, a declarative resource manager, Kubernetes CRD operators, and data-masking utilities.
A scheduling and pipeline CLI streamlining job management on the Hadoop cluster — Spark, Python, bash, arbitrary commands — adopted across data science, engineering, and analytics.
Framework exposing ML models as RESTful services over a horizontally scalable, message-based, container microservice architecture. Led the intern project team.
An NLP-based solution in R for partner and distributor name de-duplication, improving campaign and reporting data quality across the org.
A full-stack platform for capturing, querying, and replaying audit events — designed for compliance-grade traceability. End-to-end ownership: data model, API, edge delivery, and UI.
A Terraform/OpenTofu provider for consistent, convention-driven resource labeling — published to both the Terraform and OpenTofu registries. Go, provider-framework internals, and release automation.
Eight roles across four companies, current role first. Expand any role for the detail — what I built, the stack, and the impact.
Proficiency reflects depth of production use, not exposure. Bars show relative command; ▲ marks where I'm actively going deeper.
I started in business operations — automating spreadsheets with VBA because the manual process was too slow and too error-prone. That instinct, find the bottleneck and build the system that removes it, is the same one behind everything I've built since.
Most of what I know is self-taught and proven in production. I picked up the data-science stack through coursework in 2014, then spent the next decade building the real thing: ML platforms, cloud infrastructure, data systems, and the tooling that makes other engineers faster. The titles changed — engineer, lead, director — but the work stayed the same shape.
Starting on the business side means I tend to ask what a system is for before asking how to build it. As more of the engineering craft gets automated, that judgment — what's worth building, and how to make it actually pay off — is the part I lean on most.
Every project carries its own risk tolerance and appetite for using AI. For example, I built this website almost completely with AI because it is simple, can tolerate errors, and will need minimal maintenance. This is the exception to most projects I build.
While AI has changed how fast I can work, it has not changed what I'm responsible for. I use it for leverage on the parts I already understand, and I don't ship anything I can't read, reason about, and defend.