Work

What I've built, and what I learned building it.

From vector geometry in the browser to real-time support knowledge to payments compliance at scale. Here is the work, and what carried over between them.

01Case studies

Airbnb 2025 to now
  • Java
  • Kotlin
  • Kafka
  • JavaScript
  • Claude Code

Payments compliance infrastructure

Keeping a global payments platform compliant, and one step ahead of financial crime.

Backend services and Kafka pipelines that product, trust and investigations teams all lean on. Millions of hosts and guests a day, so every edge case eventually happens.

On the side: AI-assisted workflows that take a change from draft to review-ready, and the sessions that get the team using them.

LinkedIn 2024 to 2025
  • Python
  • Java
  • Kafka
  • MySQL
  • REST

Unified Content Lake

Support knowledge lived in a dozen systems. I led the pipeline that made it one.

Kafka ingestion, normalisation, and a store tuned for the bursty, read-heavy traffic of live support. Virtual assistants and knowledge hubs read from it at conversation speed.

Drove the integration across internal tools. Positive responses up 65 percent. Presented at the company-wide all-hands.

Positive response rate
+65%
Adobe 2022 to 2024
  • C++
  • JavaScript
  • Vector graphics

Quick Pen Tool, Illustrator on the Web

The pen tool is the heart of Illustrator and the hardest thing in it to learn. I led the rethink.

Core C++ on Illustrator for the Web, shipped as a public beta at Adobe MAX 2023. Quick Pen makes precise paths feel direct without loosening the geometry underneath. It is now one of Illustrator's most used tools.

Ten plus features end to end, from vector computation to GenAI text-to-vector. Spot awards two quarters running.

Quick Pen Tool
One of Illustrator's most used tools
Shipped
10+ features
Adobe 2024 · two patents
  • C++
  • Python
  • Generative models
  • Geometry

Generative vector tracing

Classic image tracing fits curves to pixels and produces paths no human would draw.

Project Vectormitate: a generative model of vector strokes, so traced output is editable, intentional geometry. Then a second method snaps those traces to the shapes the designer meant. Two patents.

02Areas of expertise

Distributed backend services

Service decomposition, idempotent APIs, retries with backoff, circuit breaking, and consistency models chosen on purpose.

Event-driven pipelines

Kafka producers and consumers at scale. Partitioning for ordering, dead-letter handling, replay and backfill.

Data modelling and retrieval

Schema design across heterogeneous sources, indexing and caching, query paths tuned for p99 under bursty load.

Systems C++ and computational geometry

Bézier curve fitting, path simplification, hit testing, and numerically stable geometry compiled for the browser.

Applied generative models

Generative vector modelling for raster-to-vector tracing, text-to-vector evaluation, and LLM-driven developer workflows.

Reliability and observability

SLO-driven alerting, structured logging and tracing, incident response and postmortems.

03Mentoring

Most of what I know about switching roles I learned the slow way.

Three companies in three years: a creative tool at Adobe, a platform team at LinkedIn, payments at Airbnb. Every switch taught me something I wish someone had just told me. So now I tell people.

Interviews, offers, or a resume that undersells you. Book a session.

Bring a specific question. "Review my resume for this role" beats "any advice?" every time.