Emre Şahin's CV (September 2022)

Emre Şahin’s CV

I have 20+ years of experience in building end-to-end software and machine learning systems, currently specializing in MLOps and Rust.

Experience

• Backend and Machine Learning Engineer at recital.software (February 2023-Current): Building ML systems focused on LegalOps. Working with Ruby on Rails backend to improve intelligent contract processing with LLMs (GPT, Claude-2 and Llama) and vector search (Pinecone, pgvector)

• MLOps Data and Pipelines Tool Development with Rust (Open Source: https://github.com/iesahin/xvc): A tool that will run data and ML pipelines in parallel, manage models and keep version history in Git, without sacrificing performance, security or ergonomics. This is my daily habit for the last two years now.

• ML Engineer and Technical Writer in iterative.ai (February 2021-June 2022): Writing documentation and creating example projects for DVC. It also involves creating katacoda scenarios, testing the new releases, dockerizing solutions, contributing to design decisions and designing a middleware server to authenticate data distribution.

• Chairman and Chief Researcher of Teknokrat Yazilim A.S. Bursa, Turkey (November 2013-Present): I founded this company to build ML solutions around my Ph.D. work. At some point we were a team of 5 engineers. We got funding for research projects about Ottoman language dictionary and OCR (dervaze), built solutions for CMU’s Human Rights Center about Arabic Video Analysis (avians), created a real time face recognition project (facebin) to be used in Makim’s access control systems. We also created mobile applications for Ottoman OCR, various web applications and cloud solutions.

• Performance Engineer and SaaS Ops Manager in DevFactory/Trilogy (January 2011-December 2011): I started this position as a remote developer to fix performance problems in various desktop software. I implemented database level optimizations and promoted to SaaS Ops Management to maintain various SaaS solutions with a 20+ person team. We were responsible to keep 6-8 software projects live with a team around the globe.

• Consultant, Project Manager, Devops and Software Developer in YD Yazilim, Ankara Turkey. (2005-April 2010): I was doing consultancy work for various custom solutions during my MS and Ph.D. studies. These involved database and backend solutions, Subversion based source code management and project management

• Teaching and Research Assistantships in Bilkent University’s Computer Engineering Department. (September 2002-February 2010): Course assistantship for all levels of undergraduate CS courses, mostly CS461 Artificial Intelligence taught by Prof. Varol Akman

• Various small programming projects during high school and CS undergraduate studies. (1995-2002): I taught myself Delphi and C and produced some usable software for notary offices at that time.

Recent Projects

• Xvc: (Solo, 2022) I’m currently building an MLOps CLI application to manage ML data dependencies, pipelines and experiments in Rust. I’m planning to release an alpha version before 2023. It will be used to version data in Git repositories, have pipelines with state machines, and great performance of the native code.

• Documentation and Get Started Projects for DVC: (Team member, 2020-22) I wrote getting started tutorials and user’s guide sections for the documentation. I built several Tensorflow and PyTorch based projects as a showcase for DVC pipelines, data management and experimentation features.

• Facebin: (Solo, 2018-19) A distributed real time face detection and recognition application in Tensorflow and PyQt. One of the servers receives ffmpeg streams from multiple cameras, identifies keyframes, serializes frames to Redis. Another server runs face detection on these key frames and stores them on Redis. A third server runs face recognition on these detected faces and stores results on Redis. The user interface is also independent from these. The system is designed to be scalable by adding consumer level GPUs as servers, or could be run on a single workstation. The recognition system uses transfer learning based on VGG-16, and has a decent performance for 100+ people with a single photo from each.

• Avians: (Solo, 2015-16): Arabic Video Analysis System runs word spotting and Arabic OCR on freely available videos. It finds key frames, detects Arabic text and searches names on these key frames. It was a continuation project for Dervaze to find missing people’s names in Syrian Civil War for CMU’s Human Rights Center. It uses DL models

• Dervaze: (Researcher & Manager, 2012-2016): An ML system for Ottoman Turkish. It has some manual transliteration between Arabic based Ottoman Turkish and Latin alphabet, synthetic data set generation, feature engineering, similarity metric experiments customized for Arabic words, experimenting with deep learning solutions, web based labeling solutions, mobile applications for Android and iOS to take a photo and get the translation (similar to Google Translate does for living languages), etc. We were successful in 19th century lithography texts but for larger adoption we needed more data and funding.

Education

• Baskent University, Department of Computer Engineering, BA (1998-2002). Highest Rank in the Department.

• Bilkent University, Department of Computer Engineering, MS (2003-2005) and PhD on analogy-making, computational semantics and Situation Theory with Artificial Intelligence (2005-2010, not finished) with Prof. Varol Akman, and PhD on ML/CV solutions for Ottoman Turkish (2012-2016, not finished) with Prof. Pinar Duygulu I completed all course work but wasn’t able to finish these due to personal complications.

Technologies

• All major Machine Learning and Computer Vision languages and technologies. (PyTorch, Tensorflow, Dlib, OpenCV…)
• All major Programming Languages (Python, C, C++, Java…) and some minor ones (Rust, Lisp, Haskell…)
• Web frameworks (Djanjo, Flask, JQuery…)
• Some Android and iOS programming (Dart, Flutter, Kotlin)
• Linux related tools and technologies for DevOps / Administration

Natural Languages

• Turkish (Native)
• English (After 30 years of use, it’s almost native)
• German (Intermediate)
• Persian (Beginner)
• Mandarin Chinese (Beginner)
• Finnish (Beginner)

Disabilities

• I have 50% hearing loss due to otosclerosis in my ears and it’s advancing. I have no problem in communications with a hearing aid but avoid speaking/listening based careers for the long term.

Personal Info

Official Name

Ibadullah Emre Sahin (he/him)

Birth

15 July 1979, Ankara Turkey

Citizenship

Turkish

Contact

i.emre.sahin@gmail.com

+90 532 261 8985