Recep Yusuf Bekci

Research

One-Warehouse Multi-Store System ⁽²⁰²¹⁾

Preprint

• Modelled one-warehouse multi-store inventory system using dynamic programming and investigated its mathematical properties
• Developed demand learning techniques to learn near-optimal control policies from censored/uncensored historical data
• Proved the optimality guarantees of proposed procedures and algorithms
• Tested the algorithms against benchmarks using a commercial dataset which has 80 million rows

Visualizing the Loss Landscape of
Actor-Critic Methods with Applications in
Inventory Optimization ⁽²⁰²⁰⁾

Arxiv

• Showed the characteristics of the actor loss function which is the essential part of the optimization
• Exploited low dimensional visualizations of the loss function and provide comparisons for loss landscapes of various algorithms
• Applied this approach to multi- store dynamic inventory control, a notoriously difficult problem in supply chain operations, and explored the shape of the loss function associated with the optimal policy

Image-based Forecasting for Fast-Fashion ⁽²⁰¹⁹⁾

PRESENTED AT INFORMS 2019/ SEATTLE

• Utilized computer vision techniques to retrieve the information from images
• Implemented transfer learning techniques to effectively use convolutional neural networks to predict product sales performances from their images
• Collaborated with Koton, hence used the data from one of the largest Fast-Fashion retail companies in Europe

Autism Prediction from Genetics ⁽²⁰¹⁹⁾

Comp 766: Graph Representation Learning Project

• We model genomic mutations in families with at least one child with Autism Spectrum Disorder (ASD) in a graph
• Utilized a heterogeneous graph neural network (GNN) variant to perform a node classification task
• Compared the GNN's performance with a gradient boosting baseline

Optimization of Customer Retention Rate in Life
Insurances Renewal and Exit Processes ^(2017-2018)

• Predicted the churn probabilities of customers using their socio-economic features and transaction history
• Used the predicted probabilities to optimize the actions and promotions to increase retention of life insurance customers via mathematical programming
• Collaborated with Aviva

Risk-Averse Airline Revenue Management with
Coherent Measures of Risk ⁽²⁰¹⁷⁾

Link for the project

• Modelled the problem using Markov Decision Processes
• Showed whether a control limit policy exists for single-leg risk-averse model with overbooking, cancellation and no-shows under risk-averse setting

Work Experience

Oracle Labs ⁽²⁰²²⁾

Axya ⁽²⁰²¹⁾

• Working on an end-to-end project that estimates production costs for metal manufacturing
• Developed an automatic feature extractor from DXF(drawing) files
• Built statistical models to effectively predict production times using generic and extracted features
• The preprint can be found here!

Aviva-Sa Insurance ^(2017-2018)

• 8 mounths project to increase customer retention
• Prediction of churn probabilities
• Optimization of promotions/actions to maximize retention

Turkish Aerospace Industries, Inc. (TAI) ⁽²⁰¹⁸⁾

• Part-time engineer at Production Planning and Control Department
• Controlled and provisioned of the master production plan
• Monitored the production process of aircraft parts and assured the implementation of the plan

ASELSAN Electronics ⁽²⁰¹⁷⁾

• Business Analist intern at IT Department
• Improved IT service desk setting using data analysis and simulation
• Analyzed the historical service data and simulated proposed new settings to compare the performance metrics

Teaching

MGSC 695: Deep Learning (McGill University) ^(2019,2020)

• Teaching Assistant
• A hands-on outlook of deep learning methods including regression, classification, convolutional neural networks, autoencoders, and forecasting time series
• Gived tutorials for Keras
• Conducted office hours, graded assignments, midterm and final projects

MGCR 472: Operations Management (McGill University) ^(2019,2020)

• Teaching Assistant
• A course on operations including strategy, process analysis, queuing systems, supply chain management, and inventory management
• Conducted office hours, graded assignments

About

Hello! My name is Recep. I graduated from Bilkent University with BSc in Industrial Engineering in 2018. I started my PhD in operation management at McGill University in the same year. My research interests are inventory control, revenue management, and reinforcement learning. I'm fortunate to be advised by Professor Mehmet Gümüş and Professor Sentao Miao. I aim to defend my thesis on Learning Applications on Inventory Control and Revenue Management by Spring 2023.

My coursework is varied in subjects operations research, machine learning, and statistics. To name a few:

• COMP 767: Reinforcement Learning (McGill University)
• COMP 766: Graph Representation Learning (McGill University)
• MPHE 740: Dynamic Optimization (HEC Montreal)
• COMP 652: Machine Learning (McGill University)
• MATH 782: Bayesian Inference,Computational Methods and Monte Carlo (McGill University)
• MPHE 743: Large-Scale Data Analysis & Decision Making (HEC Montreal)
• MATH 556: Mathematical Statistics (McGill University)
• MATH 560: Optimization (McGill University)
• MATH 523: Generalized Linear Models (McGill University)
• MATH 545: Intro to Time Series Analysis (McGill University)
• ECON 706: Machine Learning for Economics (McGill University)
• EEE 495: Statistical Learning and Data Analytics (Bilkent University)
• IE 432: Quantitative Risk Management (Bilkent University)
• IE 451: Applied Data Analysis (Bilkent University)
• CS 461: Artificial Intelligence (Bilkent University)
• IE 440: Introduction to Financial Engineering (Bilkent University)

Contact

• Twitter
• Facebook
• Instagram
• GitHub
• LinkedIN
• Mail

Elements

i = 0;

while (!deck.isInOrder()) {
    print 'Iteration ' + i;
    deck.shuffle();
    i++;
}

print 'It took ' + i + ' iterations to sort the deck.';