az✷
SENIOR SOFTWARE ENGINEER · ANDURILPERSONAL SPACE / 2026

Arthur
Zhuk.

Complex systems.
Human experiences.

I build software that holds up under real pressure. Lately that has meant upgrading MongoDB across four major versions, merging two frontends into one architecture, and taking a team from weekly releases to three a week.

Explore my work ↓View résumé ↗
FIG. 01 / CONTINUOUS EXPLORATION
SYSTEMS IN MOTION

Currently building at Anduril

BACKEND DEPTH. FRONTEND INSTINCT.SCROLL TO EXPLORE ↓
✷

A résumé tells one part of the story.

Ask a better question.

Explore my experience and how I build, or
paste a job description and get an honest fit.

01 / THE WORK

Built for the
real world.

Over a decade across defense, healthcare, SaaS, and DeFi. Different domains. The same care for how things work.

3×

Faster release cadence

Travel Syndicate Technology
55%

Fewer slow queries

Insight Rx · MongoDB modernization
10k+

Users of products I helped build

Procore · Quality & Safety
CASE STUDIESTHE PROBLEM. THE WORK. THE RESULT.
01Insight Rx

Upgrading MongoDB across four major versions

The problem
A healthcare precision-dosing platform was running on an aging MongoDB, a ten-year-old AngularJS frontend, and untyped JavaScript.
What I did
Executed the MongoDB upgrade across four major versions, then modernized the surrounding stack: JavaScript to TypeScript, and AngularJS to React.
  • 55%fewer slow queries
  • ~180fewer Sentry errors per month
  • 35%fewer P0/P1 incidents
  • 12 → 4days of cycle time
02Travel Syndicate Technology

One React architecture for Hotel, Itinerary, and Manual Booking

The problem
Legacy Ember and React code lived side by side across 6 repos and 10 engineers, with duplicated logic, weekly releases, and slow incident mitigation.
What I did
Migrated to a single React architecture, gated releases on E2E checks, improved alerting, and added PostHog analytics to see how the product was used.
  • 30%less duplicated logic across 80 modules
  • 35%fewer production regressions
  • 1× → 3×releases per week
  • 45 → 15minutes to mitigate incidents
03Procore

Letting teams ship in parallel on a Rails monolith

The problem
The Quality and Safety suite was built into a Rails monolith, with a slow Webpack pipeline and JavaScript that shipped runtime bugs.
What I did
Decoupled the monolith into micro-frontends for parallel development, tuned the Webpack pipeline, and shipped Action Plans as the first TypeScript product.
  • 20 → 5minutes of CI build time
  • 40%fewer runtime bugs in Action Plans
  • 25major features across six products
  • 500+companies using them
SELECTED EXPERIENCE2010 — PRESENT
01
AndurilCURRENTSenior Software Engineer
May 2026 - Present

Building mission-critical software as a Senior Software Engineer in a high-accountability defense technology environment.

  • Contributing to production software where reliability, system correctness, and operational clarity matter
  • Applying senior fullstack and backend engineering experience across API design, data flow, and maintainable system architecture
Backend SystemsSystem DesignTypeScriptReliability
02
Travel Syndicate TechnologySenior Software Engineer
Feb 2025 - Apr 2026

Own architecture for Hotel, Itinerary, and Manual Booking surfaces; set standards for performance budgets, state management, and release quality across 10 engineers and 6 repos.

  • Implemented PostHog analytics resulting in 1M+ tracked events in the first 3 months
  • Migrated legacy Ember + React to a unified React architecture, cutting duplicated logic by 30% across 80 modules
  • Reduced production regressions by 35% by gating releases on E2E checks
  • Improved release velocity from once per week to three times per week in the first 90 days
  • Shortened mitigation time from 45 minutes to 15 minutes via improved alerting
ReactTypeScriptPostHogE2E Testing
03
Insight RxSenior Software Engineer
Dec 2023 - Feb 2025

Led fullstack modernization for a healthcare precision dosing platform, shipping enterprise features across the entire stack.

  • Designed the Product Features module unlocking seven enterprise workflows and five new drug modules
  • Modernized JavaScript to TypeScript, reducing Sentry errors by roughly 180 per month and cutting P0/P1 incidents by 35%
  • Executed a MongoDB upgrade across four major versions, reducing slow queries by 55%
  • Replaced a 10-year AngularJS frontend with React, cutting cycle time from 12 to 4 days
ReactTypeScriptMongoDBNode.jsExpress
02 / THE BUILD

This site is
a project too.

Ask AI Arthur something hard. Here is what keeps its answers honest.

  1. 01

    One source of truth

    This page, the AI's answers, and the résumé links all read from a single profile file, so they cannot drift apart.

  2. 02

    Structured answers, not free text

    Every reply is generated against a strict schema, validated as it streams, and drawn by components I control. A malformed answer never reaches the screen.

  3. 03

    It cannot invent a link

    The model chooses from an allowlist (email, LinkedIn, GitHub, résumé). The app supplies the URL, never the model.

  4. 04

    Honest about the gaps

    It answers only from the profile, labels inference, and says when something is not there. Paste a job description and it reports where Arthur does not fit, too.

  5. 05

    Guarded and private

    Input, request size, output, and time are capped, and visitors are rate-limited per minute, per day, and site-wide, using a hashed ID. Conversations are not stored; they live in your tab.

  6. 06

    Tested without a model

    Streaming, cancellation, retry, and error paths run in offline tests against a mock model, plus a Playwright pass over the real interface.

  • Next.js
  • React 19
  • TypeScript
  • AI SDK
  • OpenAI Responses API
  • Zod
  • json-render
  • Three.js
More on GitHub ↗
03 / THE PERSON

There’s more
to the story.

Curiosity doesn’t switch off when the laptop closes. You’ll find me exploring on two wheels, on the ice, trying a new recipe, or building something just to see where it goes.

A LITTLE SOMETHING DIFFERENT

Arthur of Silver Lake Original AI soundtrack

01

The long way home.

Road biking & getting outside

02

One more experiment.

AI, blockchain & side projects

03

A competitive streak.

Hockey, basketball & gaming

04 / WHAT’S NEXT?

Good things start
with a conversation.

↗
✷   MEET AI ARTHUR