About

Updated

Currently exploring

Backend and platform engineering, distributed systems, and reliability. These are the questions guiding what I learn and build next.

Distributed systems

What if the work succeeded, but the reply disappeared?

I'm exploring how retry identity, durable state and acknowledgements fit together. A useful test is being able to explain what happens before and after each failure.

Try Request Replay

Read the note on uncertain outcomes

Platform engineering

How can one reliable boundary support several callers?

I'm interested in APIs and tools that make the safe path easy to use. That includes clear validation, useful errors and shared rules that hold across different interfaces.

Explore DeskLab's shared validation

Read about Voscus's policy boundary

Reliability

What should keep working when a dependency doesn't?

I'm learning to make failure behaviour explicit: timeouts, bounded retries, useful diagnostics and recovery. Small, reproducible failures give me something concrete to reason about.

See the boundaries in this website

Read about a narrow role for AI

A later project idea

A small durable work queue

Take the failure scenarios in Request Replay into a real backend, then make each step observable.

This is an idea for a future build. The scope would be deliberately small: submit a job, let a worker process it, and inspect what happens when delivery repeats or the worker stops.

Questions to test

  • Can a restarted worker resume without losing accepted work?
  • What prevents repeated delivery from repeating a side effect?
  • How do retry limits and backpressure keep failure from spreading?
  • Can a trace explain a job's path without exposing its payload?

I'm currently at Shopify and incoming at Stripe for summer 2027. I want to bring what I've learned about data systems, validation and backend services to infrastructure that other engineers depend on.

As I work through these questions, the smaller observations will go into Engineering notes.