About Me

Hello and salutations! My name is Behrooz Vedadian (بهروز ودادیان in Farsi) and I enjoy mathematics, arts and computer science. I currently live in Bojnord with my family. In case you’re in a hurry, here are some of the highlights of my programming career:

• Designed and implemented a C++-based MLPMulti Layer Perceptron training and inference engine (during my time at Hoda Systems, 2010)
• Developed an NIDNational Identification card reader using a crude version CNNConvolutional Neural Networks for City Bank of Egypt (2012)
• Proposed an encoder model for NMTNeural Machine Translation akin and prior to “Transformers” using GSPGraph Signal Processing techniques as my Ph.D. proposal (September, 2014; in Farsi)
• Developed the Targoman SMTStatistical Machine Translation engine as a team member (during my time at HLT Lab. 2016)
• Co-founded Targoman Co. Pjc. with my supervisor and another colleague (2018)
• Heavily contributed to development of Tarjomyar, an HATHuman Aided Translation software (during my time at Targoman. 2018)
• Developed a RESTful NMT API server on top of MarianNMT (during my time at Targoman, 2019)

Me & Academia

Born in the early days of spring in 1982, I began highscholl in 1997, a time when personal computers were rapidly gaining popularity around the world. By then I had already developed an interest in computer programming and I dreamt of creating an ASRAutomatic Speech Recognition system. Although I did not know anything about artificial intelligence, my fascination with the dream of intelligent computers was making me restless.

I started to delve into Fourier Transforms and other tools of processing audio signals. I enrolled in the Sharif University of Technology to pursue a degree in Electrical Engineering (I thought programmers must be able to provide themselves with appropriate hardware :) ). During my studies my passion for mathematics led me to immerse in signal processing. As a result, after graduation I switched to studying Communication Systems Engineering at AmirKabir University of Technology, where I earned my master’s degree.

After eight years of working in the AIArtificial Intelligence industry at Hoda Systems Co. Ltd., where I developed solutions for OCROptical Character Recognition, I felt an overwhelming desire to return to academia and pursue a Ph.D. in Natural Language Processing. This time, it was the allure of language models, which are crucial for OCR, and the captivating world of human language that drew me in. As I delved deeper into the study of language models, I became increasingly fascinated by the intricacies of human language. Although I have already unofficially abandoned my Ph.D. pursuit due to various reasons, the experience remains an important chapter in my academic journey.

Software Industry

My professional journey began as a high school student, developing small software applications for local clients in my hometown. However, it was my role as an MLMachine Learning engineer at Hoda Systems that formalized my connection to machine learning. Prior to Hoda Systems, I primarily worked with the PASCAL language, specifically Delphi, which proved advantageous as I achieved a perfect score in a programming language course at Sharif University of Technology. Additionally, in my second year at the university, I served as a teaching assistant for the same programming language course. During my time at Hoda Systems, I expanded my skills to include C++ and became proficient in writing C++ programs.

After nearly seven years at Hoda Systems, I felt the urge to continue my academic pursuits and joined the development of a national machine-translation system, funded by the ITRCIran Telecommunication Research Center. As part of the project, we redesigned the Moses Translation Toolkit to make it more maintainable and production-ready. Our team had a great time reviewing every design decision we made for the final software. The outcome of our efforts, although not all of it remains relevant today, can be found in the Targoman’s Github repositories. This project was undertaken in the HLTHuman Language Technology Lab. of the Computer Department at AmirKabir University of Technology, under the supervision of Dr. Shahram Khadivi, alongside my colleagues.

After nearly two years, I, along with my supervisor and another colleague, co-founded Targoman Co. Pjc. We went on to secure two more projects and grants, ultimately developing a neural machine translation engine for the Farsi-English language pair. At Targoman Co. Pjc, we also created an web application, which featured DLADocument Layout Analysis of PDF documents as a key component. Although I had some experience with front-end programming during my time at Hoda Systems, working on Tarjomyar was a significant challenge that allowed me to delve deeper into the realm of front-end development.

Final Words

As I look back on my journey, I am filled with gratitude and joy for the incredible experiences and opportunities that life has brought my way. Thank you for bearing with me and if you felt like contacting me you can do so by email using address vedadian at gmail dot com.