Ertug Karamatli

Email:

Summary

Self-motivated and innovative Machine Learning (AI) Engineer/Researcher/Consultant with a proven track record in machine learning, software engineering and management.

I am open to international remote work opportunities to develop machine learning systems that solve diverse real-world problems while tackling all of the challenges in the requirements gathering, data collection, model development, deployment, monitoring and automation phases. I am based in Istanbul, Turkey.

Keywords

Machine learning: PyTorch, TensorFlow, scikit-learn, Vowpal Wabbit, CUDA, MLOps, Deep Learning, Discriminative & Generative Models, Variational Autoencoders, Generative Adversarial Networks, Image & Audio & Speech & Music Processing, Recommendation Systems, Bayesian Methods, Non-negative Matrix Factorization, Source Separation, Anomaly Detection, Fraud Detection
Data science: Python, Jupyter, Spark, Pandas, NumPy, SciPy, Matplotlib, Seaborn
Programming languages & tools: Java, Python, JavaScript, Node.js, SQL, Bash/Shell, C#, C/C++, debuggers & profilers, gRPC/protobuf, Docker, Git
Web technologies: Spring MVC, Thymeleaf, JSP, JSF/PrimeFaces, Django, Flask, HTML5/CSS/JS, Bootstrap, jQuery, Angular, React, Grunt, Bower, Selenium
Java ecosystem: JVM internals & tuning, thread dump & heap dump analysis, VisualVM, Maven, Spring Framework, Spring Boot, Hibernate, JAX-RS/Jersey, AspectJ, JUnit
Hadoop ecosystem: Cloudera, HDFS, MapReduce, Hive, Impala, Spark, Flink, Flume, Kafka, Hue, Schema Registry, Parquet, Avro
Data stores & message brokers: MySQL, Vertica, SQLite, Cassandra, MongoDB, Elasticsearch, Redis, Memcached, ZooKeeper, Hazelcast, Kafka, ActiveMQ
System administration: Debian, Linux internals, scripting with Bash & Python/Ansible, Prometheus, Grafana, web & application servers, load balancers, firewalls, NAS appliances, ZFS, tcpdump/Wireshark, virtualization, Kubernetes, Google Cloud Platform (GCP), Amazon Web Services (AWS)
DevOps concepts: Continuous Integration/Delivery, Microservices, Infrastructure as Code, Monitoring & Observability

Industry Experience

Current Employment

Machine Learning Engineer
FarmLink (Canada, 100% Remote)

May 2022 – ongoing

As an independent contractor, I am working on the end-to-end development of novel machine learning applications.

The Turkish Internet Giant: sahibinden.com

I worked at sahibinden.com for 10 years. It is an online platform for classified ads (like craigslist) and shopping (like eBay). It was the most visited Turkish website and mobile app in the world. The website has a global rank of #159 on Similarweb. I left the company to complete my PhD.

AI/ML Research & Content Security Manager
sahibinden.com

Dec. 2017 – Oct. 2020

While reporting to the CTO (G. Ergül), I built and led the following two teams:
The AI/ML Research team was specialized in developing large-scale production-ready machine learning applications to improve user experience and key business metrics. Some examples of the projects:
- A vehicle recognition system (its video ad) that allows users to search for automobile ads by taking photos using the mobile app. The system predicts the make, model, production year, and body type with high accuracy and displays the matching ads.
- An ad category suggestion system that significantly reduced the time it takes to post an ad, lessened the work of the moderators that correct the ads posted under incorrect categories, and received very positive user feedback.
- A click prediction system that improves the click-through rate (CTR) of the native ads, which are similar to Google Ads. We published the results in a conference paper.
- A time-series anomaly detection system for internal use. It helps the site reliability engineering and system administration teams to notice problems early and take corrective action before it affects the users. We published the results in a conference paper.
The Content Security team was specialized in protecting the website and the mobile apps from non-human traffic such as account takeover attacks, bad bots, scrapers, and automation.
- Enriched the traffic data with internal and external data sources within the Hadoop ecosystem for ad-hoc analysis and continuous monitoring.
- Developed a rule-based anomaly detection and rate-limiting system with Flink.
- Evaluated third-party web application security products/services and developed integrations.

Software Architect
sahibinden.com

Jan. 2014 – Dec. 2017

While reporting to the Chief Software Architect (G. Demir), I became a founding member of the software architecture team that is tasked with ensuring high reliability, availability, maintainability, scalability, observability, security, and efficiency.
Took part in scaling the complex software and hardware infrastructure to cope with the peak traffic growth from 1 to 10+ Gbps over the years.
Served as a consultant to the software engineering and system administration teams by answering questions, participating in meetings, facilitating information sharing, writing documentation, and doing code reviews.
Being hands-on, spent a significant portion of my time writing code and monitoring various components of the infrastructure.
Architected the real-time bidding platform for the native ads, which are similar to Google Ads.

Senior Software Engineer
sahibinden.com

Sep. 2010 – Jan. 2014

Took part in designing and implementing a REST API for the first iteration of the mobile apps.
Took part in switching the codebase from PHP to Java.
Implemented a consistent hashing technique that allows seamless utilization of multiple Memcached servers with interoperability between PHP and Java.
Developed a credit card processing module.
Took part in architecting a modern single-page application with Angular for the “My Account” section of the website.
Developed an asset manager for JS and CSS files that integrates with Spring MVC and does automatic bundling, minification, and obfuscation.

Prior Employment

I worked at three startups as a part-time software developer for a total of 2 years while I was an undergraduate.

Software Developer
Intimatek

Apr. 2010 – Sep. 2010

Worked on a web application that automates business processes. It was very flexible/customizable and had a drag & drop UI with real-time collaboration features.
Used C# (Mono), Python, Twisted, qooxdoo and PostgreSQL.

Software Developer
Zeitin

Oct. 2008 – Apr. 2010

Took part in creating Vidi API, which was an API-oriented live video streaming service. The front-end was built using Flex/ActionScript and JavaScript. The back-end was developed with Java using Red5. The client libraries were released for easy integration.
Took part in developing a RESTful API, a load balancer for video streams, and demo applications for Vidi API.
Took part in broadcasting several events live, including TEDxReset, to hundreds of concurrent viewers.

Software Developer
Parkyeri

Jul. 2008 – Aug. 2008

Worked with several web technologies and used Java, PHP, JavaScript, and Bash.

Education

PhD in Computer Engineering
Boğaziçi University

2012 – 2022 (expected)

PhD Thesis: Audio Source Separation via Weakly-Supervised and Self-Supervised Deep Learning (Advisor: A. T. Cemgil)
We published a conference paper [4] and a journal paper [2]. Also submitted the second (final) journal paper [1], which is currently in peer review.
Passed the qualifying exam on diverse topics in machine learning.
Served as a teaching assistant for a Data Structures & Algorithms course (Instructor: M. Shalom).
To qualify for the PhD program, I also completed my MS thesis: Real-time Musical Analysis with Bayesian Methods and Its Application on an Android Tablet Computer (Advisor: A. T. Cemgil)
Selected project reports: Tempo Tracking with a Resample-Move Particle Filter, Parallel Implementation of the Bootstrap Particle Filter Using αSMC on CUDA/GPU
Coursework: Pattern Recognition / Machine Learning (E. Alpaydın), Bayesian Statistics & Machine Learning (A. T. Cemgil), Monte Carlo Methods for Scientific Computation and Data Analysis (A. T. Cemgil), Principles of Artificial Intelligence (A. A. Salah), Optimization for Big Data (F. Öztoprak & I. Birbil), Autonomous Robots (H. L. Akın), Satellite Networks (F. Alagöz).

MS in Software Engineering
Boğaziçi University

2010 – 2012

Project: Large-scale Document Similarity Search Application Using Minhash (Advisor: A. T. Cemgil)
Coursework: Principles of Software Engineering (F. Alagöz), Software Design Process (M. U. Çağlayan), Managing Software Development (İ. Oyman), Software Measures (J. M. Desharnais), Software Development as a Team (S. Üsküdarlı), Patterns in Software Engineering (A. K. Nircan), Software Quality Assurance and Reliability (O. Tosun), Software Requirements Engineering (T. Gündem).

BS in Computer Engineering
Doğuş University

2005 – 2010

Published a conference paper based on my graduation project (Résumé Information Extraction System) with my advisor (S. Akyokuş).
Built a radio-controlled wheeled robot with obstacle avoidance for an electronics project competition that had participants from various Turkish universities. Won second prize.
Participated twice in IEEEXtreme. It is a 24-hour global programming competition.
One of the founding members of the IEEE Student Branch.

Selected Publications

[1] E. Karamatlı and S. Kırbız, “Unsupervised Source Separation via Self-Supervised Training,” submitted to the IEEE Signal Processing Letters. (arXiv) (GitHub)
[2] E. Karamatlı, A. T. Cemgil, and S. Kırbız, “Audio Source Separation Using Variational Autoencoders and Weak Class Supervision,” IEEE Signal Processing Letters, 2019. (arXiv) (GitHub)
[3] İ. İşlek, E. Karamatlı, and A. T. Cemgil, “Large scale ad click prediction system,” Signal Processing and Communications Applications Conference (SIU), 2018.
[4] E. Karamatlı and A. T. Cemgil, “Tempo tracking by using a parallel particle filter on the GPU,” Signal Processing and Communications Applications Conference (SIU), 2014.

For a full publication list, please see my Google Scholar profile.

Selected Seminars

A Realistic Look at Artificial Intelligence & Machine Learning, internal seminar for all of the management including C-suite at sahibinden.com, 2020
Machine Learning at sahibinden.com, internal tech talk at sahibinden.com, 2019
Cross-Cluster & Cross-Datacenter Elasticsearch Replication at sahibinden.com, meetup talk organized by the Elastic Turkey User Group at StartersHub, 2016
Machine Learning 101, internal tech talk at sahibinden.com, 2015
Git 101 & Git Internals, internal tech talks at sahibinden.com, 2014
Scaling Live Video Streaming at Zeitin, Free Software & Linux Days at İstanbul Bilgi University, 2010

Miscellaneous

Participated in the National Artificial Intelligence Strategy Workshop organized by the Turkish government in 2020.
Completed an Engage & Grow program (12-Week Leadership, Engagement & Culture) in 2019.
Attended the Knowledge Discovery and Data Mining (KDD) 2018 Conference in London.
Attended the HP Big Data Conference 2016 in Boston.
Completed a Dale Carnegie course in 2016.
Enjoy cycling in nature.
Play electric guitar and was a founding member of Sahiband at sahibinden.com in 2014.
Was a finalist in the programming category of the Information Technology Olympiad (BITEK-O) when I was in middle school in 2002.
Have worked on various side projects since I started programming as a hobby in 2001. I also became a free/open-source software & Linux enthusiast in 2005.
My journey with computers started in front of my father’s computer when I was 4 years old in 1992. I remember playing a lot of Prince of Persia.