About Me

About Me

I am a computer scientist deeply interested in AI, Mathematics and Machine Learning. In my 18 years of experience in this field I have designed and built many solutions in a variety of industries, from manufacturing robots to AI models for banking. I actively teach what I have learned during my academic and industrial career in a number of academic and private institutions. I am also an entrepreneur and develop ideas for the future of finance.

AI

Data Science

Finance

Robotics

Experience

Work Experience

Lead Data Scientist 2018-Present

In this period I was at the Wealth Management business within the TD Bank Group. The projects I have supervised at TD are enlisted below:

  • Semantic Topic Modelling
  • Word Embedding
  • Sentiment Analysis of Customer Feedback
  • The First Canadian Investor Sentiment Index
  • Optical Character Recognition & Image Processing

Principal Data Scientist 2017-2018

My only project in Capital One Canada was to deisgn and build a transaction fraud model. This was a high-profile project in Capital One Canada/US. It consisted of an entire transaction platform on the cloud (AWS) encompassing several modules. The first use case for this platform was the transaction fraud model with a rapid re-fit and self-learning capability. I was responsible for designing, building and delivering this model and the respective self-learning pipeline for the Canadian business (went into production in the Summer of 2018). The model I have built significantly outperforms the industry standard (the Falcon model) with the ability to re-fit rapidly and self-learn. The project has been ranked among the top three most strategic projects within the entire Capital One (US/UK/Canada) for 2017 and 2018.

Senior Data Scientist 2015-2017

As a DS in Capital One UK I workd on the following projects

  • Causal Analysis of Call Volumes
  • Feature Generation Using Deep Auto Encoders
  • Application Fraud Modelling
  • Topic Modelling of Call Transcripts

Research Scientist 2010-2015

This was the period in which I started my MSc degree in Istanbul Technical University (2010) followed by a PhD in the University of Nottingham (2012). However, studying was not my only activity. I worked as a researcher in startups like ArGenit to build AI models to classify human chromosomes within images taken by powerful microscopes. I also worked as a researcher for a period of eight months at the Koc University where I used AI to generate synthetic animated characters which would show human-level gesture harmony with their speech.

Software Engineer 2003-2010

During this time, I first worked as a software engineer for an incredibly successful startup (ICAN). Later, due to my interest in robotics I shifted my focus towards PLC and manufacturing robot software design. I was exposed to concepts such as manufacturing automation, Industry 4.0, ERP systems and many industrial automation hardware and software (such as Siemens S7 PLCs, WinCC and manufacturing process management).

Education

Education

I did my PhD in the Automated Scheduling, Optimisation & Planning (ASAP) research group (School of Computer ScienceThe University of Nottingham, UK. There I was fortunate to have Prof. Ender Ozcan as my first supervisor. My thesis was focused on using machine learning algorithms in the heuristic optimization space. Though my thesis started with this paradigm, I soon came to realize that machine learning can actually be much more helpful to transfer knowledge between various (and sometimes unrelated) problem domains. In other words I became interested in algorithms which could teach each other algorithms about their expertise. This was a fruitful idea as it generated many scientific articles which were either published in high ranking journals or presented in various conferences across the globe.

I attended Istanbul Technical University (ITU) to complete my masters studies. This was a two years program where publishing scientific articles was a condition for graduation (in addition to taking courses and passing them). Although the program was titled computer engineering most of my time was spent in projects related to robotics. I have my first supervisor Prof.Sanem Sariel to thank for the opportunity. Through her projects I was exposed to a full range of robotics research such as gait generation for humanoid robots, localization and mapping, action planning, kinematics etc.
Moreover we spent two years building a team of autonomous soccer playing humanoid robots ( Aldebaran NAO) to qualify and participate in the RoboCup world challenges. Our achievements in these tournaments were significant and led to several publications (partly contributing to my masters thesis as well).
Later I was attracted to the field of heuristic optimization and continued my masters degree under the supervision of Prof. Sima Etaner-Uyar.

For my undergraduate degree I went to University of Isfahan, Iran. In Iran, back in 1998 when I entered university, a national level exam was held annualy. A student's rank in this exam determined which universities the student can go to. About 1.2M students competed for a limited number of positions within the state universities in the country. State universities are considered to have the highest level of educational quality in the country. Furthermore, the city where the university resides is also determinant in how hard it is to enter those universities. Cities like Tehran, Isfahan and Shiraz host the best universities in the country and entering them would generally require a rank of 2000 or less among all competitors. My rank in this exam was roughly 800 which allowed me to enter some universities in Tehran, more in Isfahan and almost all other cities in general. I went to the school of computer engineering and completed my degree there. I took software and hardware courses (more leaning towards software) and towards the end worked on Fuzzy Logic which attracted me at the time. My final project was an expert system which was based on this type of mathematical logic.

My Published Articles

Publications

  • S. Asta and E. Ozcan, A Tensor-based Selection Hyper-heuristic for Cross-domain Heuristic Search, Information Sciences, Vol: 299c, pp: 412--432, DOI: 10.1016/j.ins.2014.12.020. (pdf).
  • S. Asta and E. Ozcan, An Apprenticeship Learning Hyper-Heuristic for Vehicle Routing in HyFlex, IEEE Symposium Series on Evolving and Autonomous Learning Systems (IEEE SSCI - EALS 2014), Orlando, Florida, USA. pp.65--72 (pdf).
  • S. Asta, M.S. Beratoglu and A. Capar, A boundary based feature extraction method for G-banded chromosome classification, Signal Processing and Communications Applications Conference (SIU), 2012 20th , vol., no., pp.1,4, 18-20 April 2012 doi: 10.1109/SIU.2012.6204562. ( pdf).
  • Serkan Ozkul, Elif Bozkurt, Shahriar Asta, Yucel Yemez, Engin Erzin, Multimodal Analysis of Upper-Body Gestures, Facial Expressions and Speech. In proceedings of 4th International Workshop on Corpora for Research on Emotion Sentiment and Social Signals of LREC. Istanbul, Turkey May 26, 2012 (pdf).
  • E. Bozkurt, S. Asta, S. Ozkul, Y. Yemez, Engin Erzin, Multimodal Analysis of Speech Prosody and Upper Body Gestures using Hidden Semi-Markov Models. In proceeding of: IEEE International Conference on Acoustics, Speech and Signal Processing, At Vancouver (pdf).
  • Shahriar Asta, Daniel Karapetyan, Ahmed Kheiri, Ender Ozcan, and Andrew J. Parkes, Combining Monte-Carlo and Hyper-heuristic methods for the Multi-mode Resource-constrained Multi-project Scheduling Problem, Information Sciences.
  • Ender Ozcan, John H. Drake, Cevriye Altintas, and Shahriar Asta, A Self-adaptive Multimeme Memetic Algorithm Co-evolving Utility Scores to Control Genetic Operators and Their Parameter Settings, Applied Soft Computing (pdf).
  • S. Asta, E. Ozcan, A.J. Parkes, CHAMP: Creating Heuristics viA Many Parameters for Online Bin Packing, Expert Systems with Applications, doi:10.1016/j.eswa.2016.07.005 (pdf).
  • Shahriar Asta, Ender Ozcan and Tim Curtois, A Tensor-Based Hyper-Heuristic for Nurse RosteringKnowledge-Based Systems, doi:10.1016/j.knosys.2016.01.031 (pdf).
  • S. Asta and E. Ozcan, A Tensor-based Selection Hyper-heuristic for Cross-domain Heuristic Search, Information Sciences, Vol: 299c, pp: 412--432, DOI: 10.1016/j.ins.2014.12.020. (pdf).
  • R. Tyasnurita, E. Ozcan, S. Asta and R. John Improving Performance of a Hyper-heuristic Using a Multilayer Perceptron for Vehicle Routing, Proceedings of the 15th Annual Workshop on Computational Intelligence (UKCI), 7-9 Sep, Exeter, UK. (pdf).
  • S. Asta and E. Ozcan, A Tensor Analysis Improved Genetic Algorithm for Online Bin Packing, Proceedings of the Annual Conference on Genetic and Evolutionary Computation (GECCO '15), Madrid, Spain; July 2015. (pdf).
  • P. Ryser-Welch and J. F. Miller and S. Asta, Generating human-readable algorithms for the Travelling Salesman Problem using Hyper-Heuristics, Proceedings of the Annual Conference on Genetic and Evolutionary Computation - Companion (GECCO '15), Madrid, Spain; July 2015. pp.1067-1074. (pdf)
  • S. Asta and E. Ozcan, An Apprenticeship Learning Hyper-Heuristic for Vehicle Routing in HyFlex, IEEE Symposium Series on Evolving and Autonomous Learning Systems (IEEE SSCI - EALS 2014), Orlando, Florida, USA. pp.65--72 (pdf).
  • A. Yarimcam, S. Asta, E. Ozcan and A. J. Parkes, Heuristic Generation via Parameter Tuning for Online Bin Packing, IEEE Symposium Series on Evolving and Autonomous Learning Systems (IEEE SSCI - EALS 2014), Orlando, Florida, USA. pp.102--108 (pdf).
  • S. Asta, E. Ozcan, and P. O. Siebers, An Investigation on Test Driven Discrete Event Simulation, Proceedings of the 7th Operational Research Society Simulation Workshop (2014). (pdf).
  • S. Asta, E. Ozcan, A Tensor-based Approach to Nurse Rostering, Proceedings of the 10th International Conference of the Practice and Theory of Automated Timetabling, E. Ozcan, E. K. Burke, B. McCollum (Eds.), ISBN: 978-0-9929984-0-0, 2014, pp. 442--445. (pdf).
  • C. Altintas, S. Asta, E. Ozcan, T. Yigit, A self-generating memetic algorithm for examination timetabling, Proceedings of the 10th International Conference of the Practice and Theory of Automated Timetabling, E. Ozcan, E. K. Burke, B. McCollum (Eds.), ISBN: 978-0-9929984-0-0, 2014. pp. 434--437. (pdf).
  • T. Ercal, E. Ozcan, S. Asta, Soft Morphological Filter Optimization Using a Genetic Algorithm for Noise Elimination, UK Workshop on Computational Intelligence, the 14th annual series, 2014, Bradford, England. (pdf).
  • Shahriar Asta, Ender Ozcan, Andrew J. Parkes, Batched Mode Hyper-heuristics, the 7th Learning and Intelligent OptimizatioN Conference (LION13), Catania, Italy, 2013 (pdf).
  • Shahriar Asta, Ender Ozcan, Andrew J. Parkes, A. Sima Etaner-Uyar, Generalizing Hyper-heuristics via Apprenticeship Learning.EvoCOP 2013: 169-178 (pdf).
  • Shahriar Asta, Ender Ozcan, Andrew J. Parkes, Dimension reduction in the search for online bin packing policies. GECCO (Companion) 2013: 65-66 (pdf).
  • Ender Ozcan, Shahriar Asta, Cevriye Altintas, Memetic Algorithms for Cross-domain Heuristic Search. The 13th Annual Workshop on Computational Intelligence (UKCI) July 1, 2013 (pdf).
  • Shahriar Asta, A Survey on Recent Off-Shore Wind Farm Layout Optimization Methods. Technical Report, DOI: 10.13140/2.1.4524.4486, University of Nottingham, Aug 2013 (pdf).
  • Asta S, Karapetyan D, Kheiri A, Ozcan E, Parkes AJ. Combining monte-carlo and hyper-heuristic methods for the multi-mode resource-constrained multi-project scheduling problem, Proceedings of the 6th Multidisciplinary International Scheduling Conference: Theory & Applications (MISTA2013), pages 836-839.
  • Shahriar Asta and Sanem Sariel-Talay. 2011, Nature-inspired optimization for biped robot locomotion and gait planning. In Proceedings of the 2011 international conference on Applications of evolutionary computation - Volume Part II (EvoApplications'11), Cecilia Di Chio, Anthony Brabazon, Gianni A. Di Caro, Rolf Drechsler, and Muddassar Farooq (Eds.), Vol. Part II. Springer-Verlag, Berlin, Heidelberg, 434-443 ( pdf).
  • Shahriar Asta and Sanem Sariel-Talay. 2011, A Differential Steering System for Humanoid Robots,Proceedings of the 5th European Conference on Mobile Robots ECMR 2011. Orebro, Sweden. September 7,2011 ( pdf).
  • Shahriar Asta and A. Sima Etaner-Uyar, A Novel Particle Swarm Optimization Algorithm. EA'11: 10th International Conference on Artificial Evolution. Angers, France October 24, 2011 ( pdf).
  • The DataCompareR library in R which can be found on CRAN
Lectures I have Given

Teaching

Since 2013 I have been teaching in major European and Canadian universities. The topics covered during my tenure as a lecturer include (but are not limited to)

  • Machine Learning & AI (Basics)
  • Machine Learning & AI (Advanced)
  • Big Data Technologies (Hadoop, Spark, Hive, etc)
  • Cloud Technologies (AWS and Google Cloud)
  • Mathematics for Computer Science
  • Robotics
I also have been invited speaker in many conferences across Europe and North America. In addition to academic teaching, I have been invloved in teaching and career coaching in almost every institution that I have worked at. This includes guiding aspiring analysts and data scientists through their careers, establishing goals and determining milestones to achieve those goals. Currently I teach at the University of Toronto and McMaster Universities on AI, machine learning and big data.

Get in Touch

Contact