# Karim W. ElAzzouni **Senior Software Engineer** | London, UK --- ## Contact - **Location:** London, UK - **Email:** [hello@karimelazzouni.me.uk](mailto:hello@karimelazzouni.me.uk) - **LinkedIn:** [linkedin.com/in/karim-el-azzouni-29002385](https://www.linkedin.com/in/karim-el-azzouni-29002385/) - **Google Scholar:** [scholar.google.com](https://scholar.google.com/citations?user=ouy-ocQAAAAJ&hl=en) --- ## Experience ### Meta Platforms — Senior Software Engineer *London, UK | October 2021 – Present* #### Ads ML Infra Training Reliability *(July 2025 – Present)* Joined as a founding engineer of a newly formed ML infrastructure reliability team in London, focused on improving the training reliability and efficiency of massive ads ranking models trained on thousands of GPUs. The team predominantly works on Meta's [Generative Ads Recommendation Model 'GEM'](https://engineering.fb.com/2025/11/10/ml-applications/metas-generative-ads-model-gem-the-central-brain-accelerating-ads-recommendation-ai-innovation/). - Drive automated testing efforts as part of GEM Productionization to uplevel the reliability of the exploration model. #### Core Ads Quality *(December 2021 – July 2025)* Part of the central team owning the [organic bid component](https://www.facebook.com/business/help/423781975167984) in Meta's ads auction system — building cross-app and infrastructure ads delivery solutions that optimize for long-term gain. As single-threaded owner, lead, or sole contributor across key workstreams, delivered: - **0.17%** increase in ads revenue per annum - **2.5%** potential ads revenue unlock through personalization - **8.892 Engineering Years Saved (EYS)** across developer productivity projects over five years **Long-term Ads Engagement** *(January – March 2022)* Built the first version of a complex data pipeline using Meta's internal [Apache Airflow-based "Dataswarm" system](https://medium.com/analytics-and-data/i-hear-facebook-has-data-does-it-have-data-tools-a9f4c4d82ee7) to ingest ad engagement and quality metrics from multiple sources, train pre-existing ML models, and generate features and scalars for an experimental bid component aimed at improving long-term ad engagement. **Ads Conversion Familiarity "ACF"** *(March 2022 – September 2023)* Engineering owner of a new hypothesis to increase long-term revenue and user engagement by boosting familiar, high-quality ads — helping users convert on purchase ads and become recurring purchasers across Meta products. - Started as the **sole engineer**; built and launched tens of A/B tests from ideation through shipping. - Built the **first delivery infrastructure** for the initiative from scratch. - Onboarded **6 engineers** and grew the workstream to **20+ people across 3 time zones**, evolving it into a stand-alone pod. - Launched the new treatment to production, generating **0.61% iRev**. - Received strong leadership recognition for impact on Meta's ad ecosystem. **Developer Productivity and Foundations Pod** *(September 2023 – July 2025)* Co-founded the Developer Productivity and Foundation pod to promote efficiency and reliability across infrastructure — owning projects E2E from ideation through adoption. - **UQL Testing Framework "UTF":** Built an integration testing framework from scratch, integrated with [CI and Testing Infrastructure](https://asiliconvalleyinsider.com/2016/05/01/data-engineering-facebook/), and drove adoption across the team and partner teams. Saved **1.8 EYS** and prevented dozens of bugs and revenue-impacting SEVs from reaching production. - **User Features Ingestion Platform:** Built a dedicated platform enabling engineers to define and consume user features within the ads delivery system, unblocking personalization efforts for ranking alterations. Potential impact of **2.5% iRev**. Saved **2.71 EYS**. - **Experiment Monitoring, Management and Reporting Platform:** Automated experiment metric collection from multiple data sources and auto-generated standardized Google Docs reports with scheduled email delivery to stakeholders. Saved **2.7 EYS**, won **"Best Time-Saving Hack"** at the organization's Tooling Hackathon, and received Director recognition. --- ### Microsoft — Software Engineer II / Software Engineer *Copenhagen, Denmark | January 2019 – September 2021* Software Engineer at **Dynamics 365 Finance and Operations** in the **Resilient Sustainable Supply Chain (RSSC)** team, working across several platform services that other application teams across the product build upon. #### Batch Processing Service Became the **official maintainer and subject matter expert** of the Batch Processing Service — a light-weight scheduler and execution engine for asynchronous jobs across the Finance and Operations product, built in C# and X++. The service supported multiple flavors of jobs including application-team jobs, customer-authored jobs in custom deployable packages, and system jobs. - **Designed and rolled out [Batch Priority-Based Scheduling (PBS)](https://learn.microsoft.com/en-us/dynamics365/fin-ops-core/dev-itpro/sysadmin/priority-based-batch-scheduling)** — a feature that required effectively rewriting the scheduling and polling flows in the service. PBS enabled job prioritization, completely **decoupled batch from the underlying infrastructure**, and **unlocked two long-anticipated platform capabilities**: elastic compute (dynamically adding/removing batch instances on-the-fly) and zero-downtime servicing (the product's approach to blue-green deployments with near-zero disruption). - Managed backlogs, customer escalations, and acted as the central point of contact across customers, PMs, EMs, support engineers, and developers from different teams. - Authored dozens of monitoring rules in **Watchdog** (an in-house rule-based production monitoring and detection tool) for proactive monitoring of the batch service. - Contributed to **Geneva Workflows and Actions** — a low-code automation engine for production operations and incident response. #### Trials Deployment Pipeline (AxTrials2) Led the design and development of a **new deployment pipeline** for [Finance and Operations trial environments](https://dynamicspedia.com/2016/02/microsoft-dynamics-ax-trial-mode/) — a modernization overhaul of the legacy trial system, built on top of in-house deployment services and **Service Fabric** clusters. #### Cloud and Edge Validation Service Designed, implemented, and owned the **Cloud and Edge Validation Service** — an internal distributed integration testing framework enabling application teams to write and execute distributed test scenarios against environments with cloud and edge capability — a new feature allowing customers to work on portions of the application offline. #### Other Platform Contributions - Member of the **Cloud and Edge Platform** team, building platform capabilities to extend the application to support distributed workloads. - Contributed across feature management and exposure, number sequences framework, database synchronization engine, and the application's legacy kernel. - Responsible for transitioning the batch service and platform knowledge to new owners through brownbag sessions, Troubleshooting Guides (TSGs), and comprehensive documentation. --- ### Careem — Software Engineer *Dubai, UAE | April – December 2018* Software Engineer in the **Captain Payments** team, responsible for the end-to-end process of payments to captains (Careem's term for drivers), pooling trips verification, and cash trips blocking/unblocking. #### Captain Earning Service Co-built the **Captain Earning Service** from scratch — an internal web service acting as the single, reliable source of truth for all captains' transactions. - **Decreased querying time by 80+%** compared to the previous system. - Introduced **real-time captain balance calculation** and **real-time cash trips blocking/unblocking** — a feature anticipated since Careem's inception. - The service listened to events from the booking flow (monolith "core" service), computed payment splits across cash, credit card, and loyalty points, maintained debt tracking tables, and triggered block/unblock flows. - Integrated the Captain Earning Service with **Careem Solutions** (internal customer support portal), migrating all direct database access flows to go through the new service via feature toggles. #### Other Contributions - **Main contributor** to Careem's most critical microservice, **core-service** — the former monolith containing booking, pooling, routing, and other essential flows. - Maintained the cash blocking and pooling routes verification scripts. - Built a **QA bot** as a serverless **AWS Lambda** function to automate clean test data creation, simulating external transactions through pay-out vendors (e.g., Fawry in Egypt, Easypaisa in Pakistan). - Built the first version of an **API backend for the Marketplace data science team** to control their ML pipeline parameters. - Onboarded multiple new team members and cross-team contributors. --- ### Egyptian Naval Forces — Software Developer & DevOps Engineer *Egypt | October 2016 – December 2017* Full-time Software and DevOps Engineer as part of mandatory 14-month military service. #### Documents Archiving Web Application Co-designed and developed a **Ruby on Rails document archiving web application** with microservice architecture. Authored a major service driving the application and implemented several features within the main web application. Responsible for refining, maintaining, extending, and constantly testing the application to meet evolving users' demands. #### Java Middleware & Reporting - Built a **Java middleware web service** that integrated and synchronized the new application with a legacy Oracle database server — **automating a process formerly maintained manually by 3 data-entry workers**. - Implemented a **PDF report generation plug-in** (using the iText library) spanning the full report cycle: receiving user requests, relaying to the middleware, extracting data from the database, producing PDF reports, and serving them back through the main application. #### Data Center Administration Set up, administrated, and maintained a **small data center** of 6 Windows and Linux servers (physical and virtual machines) hosting web and desktop applications, a domain controller, and Kaspersky antivirus server, along with MySQL and Oracle databases. - Maintained the existing LAN network and authored incremental backup scripts. - Created and maintained a **local Git server** for team source control. - **Mentored fellow engineers** in Git usage and Linux administration. --- ### VT-MENA — Research Assistant *Alexandria, Egypt | July – October 2015* Research internship on the **"Cognitive Radio Cloud"** project — a cloud-based testbed enabling a collaborative environment for wireless communications research, with a focus on cognitive radio. - Designed, conducted, and documented **network performance benchmarking experiments** between virtual machines running on **KVM** and **VMWare** hypervisors. - Measured throughput, delay, and jitter using **iPerf**, CRUDE, and other network tools, with all experiments automated via UNIX shell scripting. --- ### VT-MENA — Software Developer Intern *Alexandria, Egypt | June – December 2014* Research internship on **indoor localization** under Prof. Moustafa Youssef, working with WiFi signals and Bluetooth Low Energy Beacons. - Built a **Wall Extraction Tool** in Java for graduate researchers to plot walls and beacon locations on floor plans, simulate intersecting signals, and calculate interference and signal strength through walls. Used Java Swing for GUI, XML for project persistence, and documented extensively with JavaDocs. - Collected WiFi signal strength data to construct fingerprint maps for indoor positioning research. --- ## Education **Faculty of Engineering, Alexandria University** — Alexandria, Egypt *(June 2016)* B.Sc. in Computer and Systems Engineering - Cumulative Grade: Very Good with Degree of Honor (81.75%) - Capstone Project: *"Learning Semantics with Deep Learning for Cross Language Information Retrieval (English-Arabic)"* - Built an integrated Cross-Lingual Information Retrieval (CLIR) framework capable of receiving Arabic queries and returning ranked relevant English documents from a parallel text corpus. - Proposed system won a **competition organized by the Egyptian Ministry of Communications**, funded with 10,000 EGP. - Processed over **5.5 million Wikipedia articles (65 GB)** across English and Arabic corpora; wrote a Python multiprocessing wrapper that reduced Arabic NLP preprocessing from days to under **90 minutes** on a high-performance server. - Research published at KDIR 2016 (see Publications). **Arab Academy for Science and Technology** — Alexandria, Egypt *(June 2011)* International General Certificate of Secondary Education (IGCSE) --- ## Publications - [**"A Genetic Algorithm-based Approach for Content Delivery in Femtocaching-Assisted Networks"**](https://ieeexplore.ieee.org/abstract/document/7925538) Michael Azmy, Karim ElAzzouni, Ahmed Abuemeira, Mustafa ElNainay. *IEEE Wireless Communications and Networking Conference (WCNC), 2017.* - [**"Using Deep Learning in Arabic-English Cross Language Information Retrieval"**](https://www.researchgate.net/publication/323884842_Using_Deep_Learning_in_Arabic-English_Cross_Language_Information_Retrieval) Omar Attia, Michael Azmy, Ahmed Abuemeira, Karim El Azzouni, Omar Hussein, Nagwa M. El-Makky, Khaled Nagy. *International Joint Conference on Knowledge Discovery and Information Retrieval (KDIR), 2016.* --- ## Skills **Languages:** C++, Python, Java, C#, X++, PHP, Rust, Hack, Ruby **Frameworks & Libraries:** ASP.NET Core, Spring Boot, Ruby on Rails, Docker, Service Fabric, Elastic Beanstalk, Terraform, Jenkins, Azure Pipelines, Elasticsearch, Kusto, Kinesis Streams, SQS. Direct expertise with Meta's internal [data tooling](https://medium.com/analytics-and-data/i-hear-facebook-has-data-does-it-have-data-tools-a9f4c4d82ee7) and [engineering infrastructure](https://asiliconvalleyinsider.com/2016/05/01/data-engineering-facebook/). **AI Agentic Tools:** Claude Code and [Meta's internal AI tools](https://ajay-arunachalam08.medium.com/meta-product-managers-use-vibe-coding-to-build-prototype-apps-for-ceo-mark-zuckerberg-544829d921b0) **Cloud:** Microsoft Azure, Amazon Web Services **Databases:** SQL Server, MySQL, Oracle, CosmosDB, DynamoDB, Redis **Big Data:** Hive, Presto, Spark