Work

SQUAKE began as carbon sold through an API: measure the emissions behind a flight, a truck route or a hotel night, buy the credits that cover them, settle with the vendor, then retire the credits on the registry. More recently we have moved up the stack. Companies book their travel through agencies, and that throws off a lot of data nobody can read; we collect it, aggregate it, and hand it back as reporting — cost per trip, patterns across teams, emissions alongside both. I lead the engineering on both halves.

The reporting side is mostly a data problem. I broke the travel-data ingestion flows into services that could actually be maintained, with clearer contracts about what gets persisted where. Then I spent a long time on the query paths behind the dashboards — collapsing aggregate queries, fixing the spend queries, making filters fast enough that people trust the numbers. I also prototyped a local-first data platform that turns messy client file drops into deterministic, reviewable pipeline runs, which is roughly where the idea for Siftbase came from.

Underneath that is the older calculation work. I re-architected the pipeline to cut memory use by 98% and stop it falling over under load, moved the carbon-query services onto RDS/Aurora to kill a class of production timeouts, and led the metrics infrastructure work that took $20k a year off the bill while making the system easier to see into.

The rest is closer to customers. I own the public Ruby SDK, the API and webhook docs, and the calculation examples — which in practice means sitting in Slack with client engineers debugging their integrations, chasing down Stripe webhook failures and checkout edge cases, and writing the status-page post-mortems when we break something.

The decision I am most pleased with was talking the team out of an AI insights feature that had grown far past what the evidence supported, and shipping a much smaller version instead. It saved months.

Ruby/Rails, PostgreSQL, Redis, Python, TypeScript, React, AWS ECS/Lambda/RDS, Terraform, CircleCI, Cloudflare Workers

Built and scaled a multi-tenant Rails application to more than four million active users across several markets, on a team spread over a lot of time zones.

The part I still tell people about: our primary payment processors were cut off with no warning. We integrated Stripe, Adyen and CoinPayments under emergency conditions and kept revenue flowing for millions of users.

Ruby/Rails, PostgreSQL, Redis, Stripe, Adyen

Built a browser-based VM deployment tool driven over WebRTC, so teams could bring up throwaway test environments on demand instead of waiting on someone else.

Also built the payment-provider integration for a neo-bank in Brazil — my first real lesson in how much of payments is edge cases.