Abdelwahab Mahmoud

Senior Software Engineer · Full-stack & Platform

I build and run production systems end to end, from React in the browser to AWS underneath. Most recently: the LLM engine and cloud platform behind an AI translation product that peaked at ~100K requests an hour.

About

I studied geophysics, then started in software in 2018 on Extreme Programming teams: pair programming, the test written first, small slices shipped continuously. Most of how I work still comes from there.

I work across the stack: React, Next.js and React Native in the product, Rails and Node.js services behind it, and the AWS platform and databases underneath. The last few years leaned toward the platform half: autoscaling, database load and LLM provider failover.

I like owning a system from the architecture decision to the incident review. When production breaks, I write down what is verified and what is only inferred before I change anything. The write-ups on this site come out of that habit.

Experience

  1. December 2024 – September 2026 1 yr 9 mo · Remote

    AI translation platform (under NDA)

    Senior Software Engineer

    Scaled AI translation to 500K jobs an hour at peak without killing a job mid-run

    The product was an AI translation platform that ran LLM jobs on AWS. I did full-stack product work in Rails and React and owned the AWS platform under it. Load sat at ~5K jobs an hour and peaked at 500K. Jobs could run for hours, and the default ECS autoscaling stopped busy tasks on scale-in, so every peak cost work in progress. I migrated the product onto ECS Fargate with Terraform and replaced the scaler with a queue-depth autoscaler on Lambda, deployment state in DynamoDB, that drains a task before stopping it. The fleet handled ~100K requests an hour at peak across background processing and LLM inference.

    • Built the multi-provider translation LLM engine with the team (Bedrock, OpenAI, Gemini): fallback chains, JSON repair and rate-limit retries kept translations flowing when a provider degraded.
    • Cleared Aurora MySQL's worst-query backlog query by query: rewritten plans, covering indexes and restructured hot reads, until CPU pressure stopped turning into incidents.
    • Fixed lock contention under concurrent writes: exclusive advisory locks on hot paths, batch-processing rewrites for load management, and connection-pool tuning after autoscaling exposed pooling contention on the shared database.
    Show 5 more
    • Built the MCP server the team's agents ran behind, exposing live databases, logs and selected APIs through authenticated, monitored access. It became the team's standard agent integration path.
    • Owned credits and billing correctness for a metered LLM product, and replaced scattered permission checks with one authorization architecture.
    • Owned an agentic triage platform that handled first-line incident and support inquiries, cutting Mean Time To Resolve.
    • Owned the GitLab CI/CD pipelines: build, e2e test, SAST and deploy gates for every service in the product.
    • Kept public APIs backwards compatible across releases with deprecation paths and contract tests.
    • Ruby on Rails
    • React
    • ECS Fargate
    • Lambda
    • Terraform
    • Aurora
    • Bedrock
    • DynamoDB
  2. December 2021 – December 2024 3 yrs · Riyadh

    Dailymealz

    Senior Software Engineer

    Pulled fulfillment out of a PHP monolith while 100K daily users kept ordering

    Dailymealz runs meal subscriptions across Saudi Arabia, and every order moves through kitchen, driver and delivery in real time. I built the order service that pushes that status to mobile and web clients over Node.js and Socket.IO, then took the fulfillment component out of the PHP/Laravel monolith behind it. A full rewrite was off the table: a team of six could not freeze features for months to run two partially consistent systems. So fulfillment came out piece by piece, as a standalone service passing state changes over AWS SQS behind a REST API contract.

    • Cut Mean Time To Resolve by 30% by instrumenting order, driver and kitchen flows with Datadog and OpenTelemetry.
    • Served hot order state from Redis in a write-behind cache with MySQL behind it, keeping status fan-out off the transactional path.
    • Migrated the real-time order service from JavaScript to TypeScript, so event and message payload mismatches failed at compile time instead of in front of customers.
    • Led a team of five building a driver location check-in system in the React Native driver app: planning, reviews and pairing.
    • React Native
    • Node.js
    • AWS
    • Datadog
    • Redis
  3. October 2020 – December 2021 1 yr 2 mo · Cairo

    Nformacy

    Full-Stack Developer

    Took a two-sided marketplace from empty repo to beta launch

    Nformacy connects business advisors with companies that need their expertise. I built it full stack in React and Ruby on Rails, from the first MVP to the beta launch, along with the cloud infrastructure and deploy pipelines it ran on.

    • Led full-stack features from requirements to deployment, and moved the web apps to a SaaS model.
    • Built mentor calendar booking on Zoom, ClickUp and calendar integrations.
    • Wrote BDD tests and owned the CI/CD pipelines.
    • React
    • Ruby on Rails
  4. January 2018 – October 2020 2 yrs 9 mo · Cairo

    DevSquads (formerly Fikr Labs)

    Software Engineer

    Built Shapa's mobile app and backend test-first, and mentored teams adopting XP

    DevSquads started as Fikr Labs, a Cairo venture studio, and became an agile consultancy that takes Extreme Programming seriously: TDD, pair programming and continuous delivery. I delivered products for its international clients. For Shapa, a US health product built around a numberless smart scale and a behavior change program, I built the mobile app and the backend behind it in React Native, Node.js and Java Spring.

    • Built software in self-organizing Extreme Programming teams with TDD, thin vertical slices and test-covered refactoring of legacy code, across React Native, Ruby on Rails, Java Spring and React.
    • Mentored teams adopting XP and Agile practices: pairing, code review discipline and iterative planning.
    • React
    • React Native
    • Node.js
    • Java Spring
    • TDD

Writing

Open source

Skills

Languages

  • PHP
  • TypeScript
  • JavaScript
  • Go
  • Ruby
  • Java (Spring)
  • Python
  • SQL

Backend

  • Ruby on Rails
  • Node.js
  • REST API design
  • service-oriented architecture
  • SQS
  • Kafka

Frontend

  • Next.js
  • React
  • Redux
  • Hotwire (Turbo & Stimulus)
  • React Native
  • Tailwind CSS

AI & LLM

  • AWS Bedrock
  • OpenAI
  • Gemini
  • agentic workflows
  • Claude Agent SDK
  • MCP
  • Langfuse
  • evals

Cloud & Infrastructure

  • AWS (ECS Fargate, Lambda, SQS, Step Functions, CloudWatch, S3, IAM)
  • Kubernetes
  • Terraform
  • Docker
  • GitHub Actions
  • GitLab CI

Observability

  • Datadog
  • OpenTelemetry
  • AWS CloudWatch
  • Langfuse tracing

Databases

  • PostgreSQL
  • MySQL
  • Aurora
  • Redis
  • DynamoDB
  • ClickHouse

Testing

  • TDD
  • Spec-driven development
  • RSpec
  • Capybara
  • Playwright
  • Cypress
  • Jest
  • VCR

Compatibility & i18n

  • backwards compatibility
  • deprecation paths
  • contract tests
  • i18n

Spoken Languages

  • English
  • Arabic (native)

Education

Faculty of Science Cairo University logo

Faculty of Science Cairo University

BSc in Geophysics

Sep 2012 – Jun 2016

Contact

Contact me

Hiring for a senior full-stack or platform role, or stuck on a production problem? Email me.

[email protected]