selected works

Agentic Infrastructure Platform

2025-present

Building a developer "platform in a box", making it easy to build your organization on Platform Engineering best principles. A single installation sets up everything application teams need to ship and run services: cloud infrastructure, Kubernetes clusters, CI/CD pipelines, GitOps, logging and alerting, all inside the organization's own cloud accounts.

Teams describe each application in one manifest file: whether it is a web service, background worker or serverless function, and which databases, caches, message queues and secrets it needs. The platform provisions those resources and generates the delivery pipeline. Every push is tested, security scanned and deployed through dev, staging and production, with approval gates and verification that the new version is running.

An AI engineering agent is built into the platform. When an alert fires, it investigates across the code, build history, logs and live systems, then prepares and tests a fix and opens a pull request for review. It also patches vulnerable dependencies, reviews pull requests and works on changes that span multiple repositories.

TypeScript, Python, Terraform, Kubernetes, Argo CD, GCP, AWS, Buildkite, Harness, Claude

Gamification Platform

2025-2026

Built the developer platform for a large enterprise's gamification initiative. It began as Platform in a Box, a portable platform that bootstraps with one terraform apply and leaves everything after to pipelines, and was then migrated onto the enterprise's live infrastructure.

App teams declare what they need in a single YAML file, smaller than a boilerplate package.json. The platform provisions all of it, and every push to main runs the full path to production: tests, build, security scans, infrastructure provisioning, deployment and end-to-end tests. Teams ship without filing tickets or waiting on other teams.

The core challenge was preserving that contract while the infrastructure underneath it was replaced. The original GCP projects, clusters and registry were swapped for the enterprise's shared multi-tenant clusters, Istio mesh, mandatory image registry, policy gates and security scanners. The enterprise keeps clusters, service identities and container images in separate GCP projects; the platform wires each service across those boundaries, so app teams never deal with them. The migration ran with zero downtime on both sides, and apps already on the original platform moved over without manifest changes.

The platform hosts the product's Kotlin backend services, a Node.js BFF and a browser game. The product launched in June 2026, and in its first week the platform served 5.8 million requests at a 99.986% success rate across roughly 97,000 game sessions from nearly 90,000 customers.

Terraform, GCP, Kubernetes, Istio, Harness, ArgoCD, Helm, Cloud SQL, Python, FastAPI

Robotics Data Platform

2024-2025

Led architecture of a domain-driven data platform for autonomous welding robots. The client's robots use 7-DoF arms with arc-welders and computer vision, generating massive amounts of data from laser scanners that produce point clouds and 3D models.

The core problem: as the company scaled, data management evolved organically into tightly coupled "scraper-style" pipelines. Teams faced a false choice between brittle systems that break on schema changes, or slow systems requiring cross-team coordination for every change. AI and robotics experts were spending more time debugging pipelines than advancing core technologies.

The solution was a platform organized around domains with clear ownership, bilateral API contracts (consumer-specific APIs rather than one shared API), and abstraction layers freeing domain owners from infrastructure concerns. Consumers authenticate once, then get route-level access approved by each domain's owner. The platform includes an edge proxy with write buffering and read caching for offline factory floor operation, syncing to cloud when bandwidth allows.

Data now reaches AI model training and robot operations in real time, replacing overnight batch jobs that shipped it to the cloud. Mean time to discover and resolve incidents dropped by 80%, and the platform is designed for a 100x increase in data volume.

Delivered via 3-week POC, 6-month build, and 3-month extension. The platform now powers production robotics operations. Read the case study.

Python, FastAPI, PostgreSQL, RabbitMQ, Terraform, AWS, GitHub Actions

Wendy's Digital

2023-2024

Full stack engineer on the web ordering platform for Wendy's, one of the world's largest quick service restaurant chains with 7,000+ locations globally.

Served as liaison between the web development and platform engineering teams during AWS to GCP migration. Implemented a dual-repository sync pattern where the AWS repo remained source of truth for application code while the GCP repo maintained platform-specific configurations. The migration used single-image multi-environment deployments with runtime configuration injection, ArgoCD for GitOps, and template-based Nginx configuration.

Contributed to FreshAI, Wendy's AI-driven drive-thru ordering system built with Google Cloud. FreshAI processes tens of thousands of orders daily across hundreds of restaurants with 99% accuracy.

Kickstarted a modern Next.js 14 rewrite of the 318K-line ordering application, introducing Server Components, Zustand state management, and domain-driven architecture.

React, Next.js 14, TypeScript, Redux, Zustand, Docker, Kubernetes, ArgoCD, GitLab CI, GCP


Other Projects