Loading...
⌨️ Typing Speed Challenge
0 WPM 100% 0.0s

Hey, I'm

Mario Khoury

I build |

Software Engineer with 5+ years building high-throughput distributed systems at Amazon scale (800+ warehouses, 4 regions). IEEE-published researcher in computer vision. I design backend services that handle millions of daily transactions and architect monolith-to-microservices migrations with zero downtime.

View My Work Download CV
Java Spring Boot AWS Microservices DynamoDB
Mario Khoury — Software Engineer
0+ Years Exp.
0+ Projects
0+ Competitions
Scroll Down

01. About Me

I wrote my first program at 14 — a Hangman game on a Casio calculator during math class. That spark turned into a career that's taken me from Beirut to Munich to Luxembourg, building everything from mission-critical warehouse systems serving 800+ fulfillment centers to synthetic datasets with NVIDIA that pushed the boundaries of industrial computer vision.

When I'm not coding, I'm probably on a bike somewhere in Europe. I've cycled from Croatia through Bosnia and Slovenia into Italy, and from Portugal across Spain. I'm also a regular hiker — Olperer Hütte and Seebensee in the Austrian Alps are hard to beat.

On the technical side, I specialize in system design — from high-level architecture and API contracts to low-level data modeling and migration strategies. I've authored 6+ design documents reviewed with principal engineers, covering systems that handle millions of daily transactions across multiple regions.

Luxembourg City, Luxembourg
French, English, Arabic
Cycling, Hiking, Gym, Chess

Fun Facts

0 Countries Cycled
0 Hikes Completed
0 Countries Visited
0 Hackathons Attended

02. Experience

Software Development Engineer (SDE I → SDE II)

Amazon

Oct 2022 — Present
  • Own a mission-critical Warehouse Management System (Java/Spring Boot) serving 800+ fulfillment centers across 4 regions (EU, US, CN, JP), processing millions of daily inventory transactions.
  • Diagnosed and resolved a critical concurrency bug in SQS message processing that caused cascading thread locks — eliminated 100% of recurring manual retries and restored system stability across all regions.
  • Designed, built, and shipped a multi-box inventory storage feature — estimated $60–70M annual business impact — enabling multi-carton handling across 800+ warehouses.
  • Engineered configurable DPI printing for shipping labels, reducing new warehouse launch time by 75% (1 month → 1 week) and cutting operational costs by $2M+ annually.
  • Led system design (HLD + LLD) of 2 core microservices (Unload & Receive, Task Orchestration) for Amazon's WMS 2.0 platform — architecting the monolith-to-microservices migration using ECS Fargate, DynamoDB, API Gateway, and event-driven architecture (SNS/SQS).
  • Architected a hybrid migration strategy enabling gradual warehouse cutover from monolith to microservices with zero downtime and zero data loss.
  • Collaborated cross-functionally with Packing and Transportation engineering teams to align API contracts and system integrations with real-world logistics workflows.
  • Mentored junior SDEs through onboarding and first project delivery — led code reviews and design sessions to accelerate ramp-up.
Java Spring Boot AWS Microservices SQS/SNS Lambda ECS Fargate DynamoDB API Gateway System Design CI/CD

Computer Vision Research Intern → PhD Candidate

BMW Group

Jan 2021 — Jun 2022
  • Co-created SORDI.AI with NVIDIA — an open synthetic dataset of 200,000+ photo-realistic annotated images for industrial object detection, adopted by researchers and manufacturers globally.
  • Co-authored the "Synthetic Object Recognition Dataset for Industries" paper (SIBGRAPI 2022, IEEE), advancing research in synthetic-to-real transfer learning for industrial applications.
  • Achieved up to 99% detection accuracy on real-world industrial assets using synthetic-only training data, validating synthetic-to-real transfer learning at production scale.
  • Eliminated thousands of hours of manual annotation by replacing human labeling with auto-labeled synthetic datasets, reducing data preparation costs by an estimated 80%.
  • Built scalable data pipelines with NVIDIA Omniverse to generate, clean, and evaluate hybrid datasets for production deployment.
Python TensorFlow PyTorch Docker NVIDIA Omniverse Computer Vision

Backend Engineer

FOO (Fintech)

May 2020 — Mar 2021
  • Joined as an intern and progressed to part-time backend engineer — contributed to a Visa-partnered digital wallet & payment platform serving the MENA region.
  • Increased unit test coverage from 10% to 90% (9x improvement) using JUnit and Mockito, reducing production incidents by an estimated 60%.
  • Built and maintained RESTful APIs (Spring Boot, PostgreSQL) powering digital wallet services — onboarding, virtual card issuance, QR payments, and P2P transfers — serving 10,000+ daily active users.
  • Implemented SonarQube code quality monitoring and load testing with JMeter, improving code maintainability and identifying performance bottlenecks before production.
Java Spring Boot REST APIs PostgreSQL JUnit Mockito SonarQube JMeter Git

Android Developer (Freelance)

Interpreti (Startup)

Jan 2020 — Dec 2020
  • Developed the Android application for Interpreti — an "Uber for interpreters" platform offering on-demand, real-time interpretation for business meetings across countries.
  • Built key features including user authentication, real-time synchronization, push notifications, and Firebase database integration.
  • Integrated Sinch API to enable Wi-Fi-based voice and video calling between clients and interpreters.
  • Collaborated in a 4-person team (2 Android, 2 iOS) to deliver cross-platform mobile apps and a companion web platform.
Java Android Studio Firebase Sinch API REST APIs

Web Developer Intern

Berytech

Jul 2015 — Oct 2015
  • Built a full-stack website from scratch as capstone project at Beirut's leading tech hub — early hands-on experience with web architecture, databases, and front-end development.
HTML CSS JavaScript SQL

03. System Design

I've authored 6+ high-level and low-level design documents, all reviewed with senior SDEs and principal engineers. Here are the architectural patterns I've designed for production systems — described at a public-safe abstraction level.

Monolith → Microservices Migration

Architected the decomposition of a 15-year-old monolithic WMS into 10+ domain-bounded microservices. Designed a dual-run migration strategy where old and new systems operate in parallel — enabling gradual warehouse cutover with zero downtime while maintaining feature parity across both systems.

Scale 800+ facilities, 4 regions, migration ongoing
Pattern Strangler Fig, Dual-Run, Domain-Driven Design
Challenges Concurrency issues between old/new systems, feature parity validation, dual-system consistency
Design Docs Multiple HLD + LLD documents reviewed with Principal Engineers
ECS FargateAPI GatewayDynamoDBSNS/SQSLambda

Event-Driven Task & Work Orchestration

Designed the full task and work lifecycle — from creation through assignment, execution, and completion — using event-driven architecture across 200+ interconnected components. The core challenge was mapping dependencies between services and ensuring reliable event delivery at scale.

Architecture 200+ components (SQS, Lambda, DynamoDB, SNS, ECS, Kinesis, EventBridge)
Consumers 10+ downstream services subscribing to lifecycle events
Pattern Event Sourcing, Pub/Sub, DLQ for failure isolation, idempotent consumers
Key Decision Chose choreography over orchestration to reduce single points of failure across 200+ components
DynamoDB StreamsKinesisLambdaSNSSQSEventBridge

Multi-Tenant REST API Platform

Designed RESTful API services supporting 3 tenants across 800+ warehouses with tenant-scoped data isolation. Co-designed the domain-based sharding strategy, sizing shard counts based on expected requests/sec and data volume per domain. Implemented API versioning for backward-compatible schema evolution.

Scale 3 tenants, 800+ warehouses, millions of daily transactions
Pattern Multi-Tenancy, Domain Sharding, Optimistic Concurrency
Key Decisions Shard count based on req/sec + data volume, composite partition keys, idempotency with TTL, If-Match versioning
API Surface 14 REST operations, OpenAPI spec from Smithy models
JavaAPI GatewayDynamoDBECS FargateSmithy

04. Projects

Robotics & IoT

Spy Robot

University team project (scored 18/20). I handled the AI and Raspberry Pi programming — implemented A* pathfinding so the robot could navigate from point A to B while avoiding obstacles in real time. The hardest part was getting collision avoidance to work reliably with the pathfinding algorithm. Includes live camera streaming for remote monitoring.

PythonRaspberry PiA*Computer Vision
Automation

Instagram Bot

A personal challenge: how many followers can a fully automated bot grow from zero? Built auto-liking, commenting, following, saved post downloading, and scheduled uploads at randomized times. Result: 3K followers and 248 posts on @smirqyp — all without touching the app.

PythonSeleniumAutomation
Automation

Infinite Craft Bot

I loved the game's concept but wanted to see how far a bot could push it. Built an automated player that combines elements, discovers new ones, and persists everything to a database. Left it running overnight — woke up to 10,000+ unique crafts discovered.

PythonSeleniumWeb ScrapingAutomation
Automation

Lebanese Currency Tracker

Built out of necessity. During Lebanon's economic collapse (the currency crashed from 1,500 to 110,000 LL/$), I was living in Germany and needed to track exchange rates to know when to send money home. Scrapes live rates and sends email + WhatsApp alerts on price swings.

PythonSeleniumWeb ScrapingAutomation
Automation

Chrome Dinosaur Bot

Plays Chrome's offline dinosaur game autonomously using pixel recognition to detect obstacles and trigger jumps/ducks in real time via OpenCV.

PythonOpenCVSeleniumAutomation
Data Science

WhatsApp Chat Review

Built for fun — I like numbers. Analyzes exported WhatsApp group chats to find out who messages the most, the most-used words (turns out "yalla" dominates every Lebanese group chat), emoji stats, link counts, and activity patterns. Visual dashboards make the data easy to explore.

PythonpandasNumPyMatplotlibSeaborn
Automation

Amazon Price Tracker

Monitors product prices on Amazon and sends email alerts when a price drops below your target threshold. Set it and forget it.

PythonSeleniumWeb ScrapingAutomation
Game Dev

Conway's Game of Life

Interactive Pygame implementation of Conway's Game of Life with cell toggling, adjustable simulation speed, and save/load game state functionality.

PythonPygame
Automation

10 Fast Fingers Bot

Typing game bot that uses OCR to read words from the screen and types them at a configurable words-per-second rate. Set your target WPM and watch it go.

PythonSeleniumOCRAutomation
Automation

Cookie Clicker Bot

Automated bot that plays Cookie Clicker — clicks the cookie, calculates the most cost-efficient upgrades, and buys buffs automatically for optimal progression.

PythonSeleniumAutomation
AI & Machine Learning

Titanic Survival Prediction

Kaggle classification project predicting passenger survival on the Titanic using feature engineering and scikit-learn models.

Pythonscikit-learnpandasMatplotlib
AI & Machine Learning

House Prices Prediction

Kaggle regression project predicting house prices based on property features using scikit-learn with data preprocessing and feature selection.

Pythonscikit-learnpandasSeaborn
AI & Machine Learning

Diabetes Prediction

Classification model predicting diabetes based on clinical features (glucose, blood pressure, insulin, BMI, age) using scikit-learn.

Pythonscikit-learnpandasMatplotlib
AI & Machine Learning

Email Spam Classifier

NLP classification model that detects spam emails using text preprocessing and scikit-learn classifiers.

Pythonscikit-learnpandasNLP
Web Development

Portfolio Website

This site — built from scratch with zero frameworks. Features an RPG game mode with a full explorable world, dark/light themes, 3D tilt cards, a command palette (Cmd+K), typing speed challenge, Matrix rain effect, and a Konami code easter egg. Because why not.

HTMLCSSJavaScript

05. Skills

Java

Expert · 7 years

Python

Expert · 5 years

TypeScript

Intermediate

JavaScript

Intermediate

C#

Beginner

C++

Beginner

SQL

Advanced · 5 years

HTML / CSS

Intermediate

Spring Boot

Expert · 5 years

JUnit / Mockito

Expert

TensorFlow

Intermediate

PyTorch

Intermediate

scikit-learn / pandas / NumPy

Intermediate

Firebase

Intermediate

AWS

Advanced · 3+ years

Docker

Intermediate

Git

Expert

CI/CD Pipelines

Intermediate

Linux

Intermediate

CloudFormation / IaC

Intermediate

DynamoDB / PostgreSQL / MySQL

Advanced · 4 years

IntelliJ / VS Code / Kiro

Expert
System Design
Design Patterns
Microservices
AI & Machine Learning
Cloud Computing
Computer Vision
Web Scraping & Automation
REST APIs
Data Structures & Algorithms
Agile / Scrum
Event-Driven Architecture
CI/CD & DevOps
Distributed Systems
Testing (TDD / Integration / Load)

06. Hackathons & Competitions

BMW Hackathon — Machine Learning Track

1st Place (ML Track) · 4th Overall out of 40 Teams

Solved a real BMW problem: filtering thousands of daily internal requests to classify relevance for implementation. Built a classification model for request relevance and a regression model to estimate financial impact. Delivered a working ML prototype in under 48 hours, recognized for technical robustness and business value.

Pythonscikit-learnpandasFlask
Lebanon, 2019

Beirut AI Hackathon

1st Place out of 50+ Teams

Built a recommender system leveraging a real grocery store dataset to suggest local substitutes for imported products. Delivered a prototype and demo video with a team of 4 (iOS, Android, Backend, AI), presenting a solution to improve Lebanon's food independence. Recognized for practical impact and scalability.

PythonRecommender SystemsBig Data
Remote, 2020

MIT Lebanon Challenge

3rd Place among 1,000+ Participants

Built a prototype platform enabling consumers to order directly from farmers, eliminating exploitative middlemen who inflated prices by 5–10x. Designed a delivery model where farmers set daily community pick-up points, reducing logistics complexity. Recognized for innovation and execution speed.

JavaSpring BootWeb App
Remote, 2020

Google Competitions (Hash Code & Kick Start)

Top 5 in Lebanon — Hash Code

Competed in 2 Hash Code and 2 Kick Start editions (2020–2021). Hash Code challenges included optimizing traffic light schedules to reduce city congestion, and maximizing scanned book value by selecting optimal library ordering under scanning constraints. Ranked 5th in Lebanon.

JavaAlgorithmsData StructuresOptimization
Remote, 2020–2021

07. Education

Master (M1) in Software Engineering

Saint Joseph University of Beirut

2020 — 2021 · Beirut, Lebanon

Software Engineering, Data Structures & Algorithms, Design Patterns, and Big Data.

Design Patterns Big Data Algorithms

Bachelor's in Computer & Communication Engineering

Saint Joseph University of Beirut

2017 — 2020 · Beirut, Lebanon

Core curriculum in programming languages, mathematics, physics, and communication systems.

Programming Mathematics Physics

French Baccalaureate in General Sciences

Collège Notre-Dame de Jamhour

2003 — 2016 · Beirut, Lebanon

General science track with a strong emphasis on mathematics and scientific reasoning.

General Sciences Mathematics

08. Publications

Synthetic Object Recognition Dataset for Industries

Mario Abou Akar, Fady Alnajjar, et al.

SIBGRAPI 2022 — 35th Conference on Graphics, Patterns and Images (IEEE)

Co-authored with NVIDIA and BMW Group. Introduced SORDI.AI — a synthetic dataset of 200,000+ photo-realistic annotated industrial images for object detection. Demonstrated that models trained on synthetic data achieve up to 99% accuracy on real-world industrial assets, validating synthetic-to-real transfer learning for warehouse and factory environments.

Computer Vision Synthetic Data Object Detection Transfer Learning NVIDIA Omniverse
Read Paper SORDI.AI

I also write on Medium

Technical articles on backend engineering, system design, and lessons from building at scale.

Read Articles

09. Get In Touch

Say hi

Whether it's a project idea, a hackathon team, or just a conversation about cycling routes in Europe — I'm always happy to connect.

Currently open to Senior Software Engineer and Backend Engineer roles in Europe (Luxembourg, Germany) or remote.

Email mariokhoury23798@gmail.com
LinkedIn linkedin.com/in/mariokhoury4
GitHub github.com/mariokhoury4
0 / 2,000