Skip to content
View ziadalzarka's full-sized avatar
🏠
Working from home
🏠
Working from home

Block or report ziadalzarka

Block user

Prevent this user from interacting with your repositories and sending you notifications. Learn more about blocking users.

You must be logged in to block users.

Content in all repositories owned by your account will be closed.
Maximum 250 characters. Please don’t include any personal information such as legal names or email addresses. This note will only be visible to you.
Report abuse

Contact GitHub support about this user’s behavior. Learn more about reporting abuse.

Report abuse
ziadalzarka/README.md

Ziad Alzarka

Senior software engineer and systems architect. 10 years building data-intensive platforms in two domains where a wrong number is somebody's real problem — finance (double-entry ledgers, reconciliation, AP automation) and medicine (AI diagnosis, claims, lab data pipelines).

Lately a lot of my work is AI in engineering: I set my team's standards for AI tooling, and built AI reviewers that have checked 1,700+ pull requests, caught 45+ bugs and stopped 5+ security vulnerabilities before merge.

ziadalzarka.dev · Résumé · LinkedIn · Medium · ziadalzarka@gmail.com


What I work on

Distributed & offline-first Sync engines, replication topologies, conflict resolution, clock design — and reconstructing state from event logs when a datastore is gone
Fintech data Payment-provider integrations, transaction matching, double-entry ledger design, and the pipelines underneath them
AI in the engineering loop AI code and security reviewers, ticket-to-PR agent workflows, MCP servers, and fitness functions that keep AI-written code consistent
Architecture & product Zero-to-production builds and second opinions. Kotlin/JVM, TypeScript, React, NestJS, Postgres, AWS, GCP

Experience

Light — Senior Software Engineer, Procure-to-Pay & Record-to-Report · Copenhagen, Denmark (remote) · Dec 2025 – present
Led intercompany invoicing end to end and built three-way matching. Root-caused and fixed three production incidents, including a platform outage; cut failed bank payments from ~3% to near zero and made the integration test suite ~10x faster. Built the team's AI code and security reviewers and Lucid, its docs and architecture decision record system.

Ankor.app — Senior Software Engineer · Malmö, Sweden (remote) · Jul – Nov 2025
Led a financial reconciliation platform from inception to production — millions of transactions a month at 98%+ reconciliation accuracy across 5+ payment providers, with dbt pipelines behind transaction matching and automated double-entry journals.

Adia Health (formerly Flow Health) — Senior Software Engineer · Los Angeles, US (remote) · Jan 2021 – Jun 2025
Led development of an AI medical assistant for laboratories, physicians and insurers. Recovered a completely lost production database of ~1.5 billion records by replaying events out of Kafka, CloudWatch and S3, moved Cassandra to DynamoDB for 90% lower cost, and cut latency by 35% and a further 20%.

Earlier, all in Cairo, Egypt:

  • CoatConnect · 2019 – 2020 — developed a medical events platform, refactored it end to end and got it to a 1.7s Largest Contentful Paint.
  • BirdCloud · 2017 – 2018 — sped up delivery with a reusable component library and ERP modules, growing client capacity 20% and revenue 30%.
  • Magmooty (formerly Coretrix) · 2016 – present — a platform for tutoring centres, later rebuilt as an offline-first Rust app.

Things I've built

Kin.app (source private) — a bill-splitting app on the App Store and Google Play that splits per item instead of per head, so nobody pays for a coffee they didn't drink. Designed, built and shipped solo. An AI receipt reader at about 0.8¢ a scan, a voice assistant that turns live speech into receipts, an MCP server for personal AI agents, and a double-entry ledger under every balance. 150+ fitness functions fail the build when a change breaks an architecture rule, which is what makes "written with coding agents, reviewed line by line" mean something.

peel — a TUI diff reviewer that stages what you just reviewed. Every local diff tool is read-only, so reviewing and git add end up as two passes over the same diff. peel is one pass: read a file, press s, and it's staged, folded away, and the next file is in front of you.

brew install ziadalzarka/tap/peel

icloud-image-labeler — auto-labels an Apple Photos library with any OpenAI-compatible LLM, writing keywords, titles, descriptions and OCR text back into Photos.app. Built for a local model in LM Studio, so the photos never have to leave the machine.

Writing

Talks and videos


I started coding at 12 and took my first paid job at 16. Most of what I write at work can't be shared publicly, so this profile shows only a small part of it — the full picture is at ziadalzarka.dev.

If you have a system that has to be right and currently isn't, get in touch.

Pinned Loading

  1. peel peel Public

    A terminal diff reviewer that stages what you just reviewed

    Go 1

  2. icloud-image-labeler icloud-image-labeler Public

    Auto-labels your Apple Photos library with a local LLM — keywords, titles, descriptions, and OCR text

    Python