Doctoral research in physics, an MBA, ten years of engineering at Uber and HERE Technologies. But what actually matters is what got built. From an early solo contract to navigation infrastructure serving all of Uber, the through-line is the caliber of the work — not the size of the logo. First-person, no sales copy.
Kitco Metals — Data Consultant (Contract), 2009
MapReduce Before It Had a Name, for a Gold-Price Website
The problem
Kitco — an established precious-metals company running one of the web's most-trafficked gold-price sites — needed a distributed log-processing pipeline. 20+ GB datasets shipped in on physical DVDs, no cloud, no Hadoop yet
Scope
One-person contract. Bash-script parallelism across a handful of machines — patterns pulled from scientific computing, arrived at independently of Google's 2004 MapReduce paper
What shipped: A polynomial regression model predicting gold-price fluctuations — two days of data forecasting two days out, R² ~90%. My first paid engineering contract, and the same instinct for right-sized infrastructure I still bring to much larger problems: match the tool to the actual scale of the problem in front of you.
Bash
Distributed Processing
Regression Modeling
Solo Contract
Uber — Staff Software Engineer
Navigation Platform at 1,400+ Nodes Per Data Center
The problem
Off-route detection algorithms needed continuous improvement without risking regressions across one of Uber's largest services
Scope
Serving all Uber apps globally. Java core, on-device iOS/Android binaries, Python simulation infrastructure
What shipped: Python-based parallel simulation infrastructure for safe algorithm iteration. On-device software for real-time cloud data ingestion. GDPR-compliant data deletion across distributed databases spanning two clouds, as part of the Postmates acquisition integration.
Python
Java
Distributed Systems
GDPR Compliance
Go
HERE Technologies — Engineering Manager
Autonomous Driving Capture Team, 2 to 14 Engineers
The problem
Build a team and capture system capable of powering the next generation of HD mapping vehicles for autonomous driving
Scope
New-generation HERE True capture vehicles, small-scale collection systems, full hardware-software integration
What shipped: Grew the Capture Systems Software team from 2 to 14 engineers. Designed and delivered the capture architecture that shipped in production vehicles. Built real-time and batch analytics for the global mapping fleet using Apache Storm.
Team Building
Apache Storm
Python
Streaming Analytics
Hardware-Software
Postmates / Uber — Senior Engineer II
Python 2 to 3 Migration Across a Full Monorepo
The problem
Migrate the entire Postmates monorepo from Python 2 to Python 3 without breaking critical delivery infrastructure
Scope
Core infrastructure monorepo, distributed testing across Kubernetes pods, Uber's wider Python service footprint
What shipped: Completed the full Python 2-to-3 migration at Postmates. At Uber, scraped and analyzed the dependency graph for Python microservices, targeted the top ~200 libraries covering roughly 93% of services, and drove the company-wide migration — including an overhaul of the open-source tchannel-python library.
Python
Kubernetes
Dependency Analysis
Open Source
Westover Labs — Founder / CTO
An AI Agent Team That Ships Code Around the Clock
The problem
Build a solo-founder development operation that could produce team-level output across iOS, backend, and infrastructure simultaneously
Scope
Two App Store apps, two FastAPI inference backends, ONNX runtime serving, full CI/CD, self-hosted infrastructure
What shipped: A 12-persona AI agent team (Familiar) operating as a virtual engineering org — planning, coding, reviewing, and deploying continuously. Two iOS apps live on the App Store: Ghost Hunter (paranormal detection) and SkinVault (privacy-first photo classification), both running inference on-device. Public technical talk at LSST 2026 Berlin on AI-augmented development for scientific infrastructure.
Claude AI
ONNX Runtime
FastAPI
React Native
Ansible
Cloudflare