• Developed an end-to-end framework that generates client insights such as customer journeys, supply/demand analyses, behavior-based segmentation, trend tracking, and retention in real-time leveraging supervised/unsupervised machine learning techniques• Created application monitoring processes for error detection/resolution and metrics collection to ensure model performance consistency

• Developed and operationalized a text-based recommender system that drives client engagement leveraging advanced text analytics and Machine Learning classifiers• Productionized a document parser by coordinating with application teams to ensure seamless integration with other internal systems• Formed the technical interview process and mentored interns/junior team members

• Built a document parser for rule validation as part of a team using NLTK and gensim for rule based phrase extraction and similarity measurements, and identification of actionable elements of sentences using Question-Answer based Semantic Role Labeling• Implemented/tested topic modeling algorithms such as Hierarchical LDA, Latent Feature Topic Modeling, and Dirichlet Mixture Models, and self taught clustering with a Convolutional Neural Network to obtain document representations in order to improve performance of a recommendation system• Built a resume parser as part of a team for specific feature extraction using key phrases and Regular Expression Parser/Tagger available in NLTK

Worked as part of the Enterprise Systems Optimization Lab at UIUC. My work included automation of data extraction processes, designing a database in AWS RDS for storage, and implementation of Machine Learning models for customer suggestion mining.• Automated the data extraction process from amazon.com for product details and customer reviews by using the Selenium framework and the Python Scrapy package• Designed a program on a Python platform to enable product search given any product keyword by the user; aim of the research is to find a correlation between the rating and sales rank of an item and sentiment of the reviews• Created a MySQL database on AWS RDS to load the extracted data in order to ease access at the analysis stage; this database will be available to public in the future• Built an NLP pipeline using tools such as Stanford CoreNLP, NLTK and Regular Expressions for normalization (stemming/lemmatization), POS tagging and dependency parsing • Implemented/tested deep learning architectures such as CNN/LSTM for automated suggestion detection using Keras; training dataset is based on grammatical rules, extracted frequent patterns from suggestion forums using a Python implementation of PrefixSpan, and topics existing in customer reviews identified by a partially labeled LDA model

• Built and deployed an image scraper on EC2 using the Scrapy package in Python; the scraper crawls four automobile websites and filters out non-car images using YOLO object detection, sends images of completed jobs to a designated S3 bucket

• Developed an algorithm on a Java platform that analyzes and groups historical data of yearly client transactions based on correlation• Conducted statistical analysis to determine probabilities of single & multiple items being ordered using Minitab and the Java algorithm to create a benchmark for the simulation stage • Assigned new warehouse locations of commonly ordered items based on the probability of being ordered and rectilinear distance to the shipping area• Created a test set of orders using probabilities to validate simulation outputs of daily warehouse processes with present and new item locations• Designed an algorithm using Java to simulate the sets of orders created using the variance reduction technique• Achieved a 50% decrease in daily process times with the new warehouse design• Prepared client facing documents including reports, decks and software• Client interactions on a weekly basis in an agile manner to update goal, progress of project and deliverables

Information Technology services department of Eczacibasi Group, a leading industrial group of companies founded in Turkey• Analyzed current duties of the department’s database administrators and improved the existing duties by making additions in accordance with the contemporary needs• Conducted research on latest technology trends and recommended new services that the IT department could provide to customers

Airline IT solutions provider to all travel industry companies• Designed algorithms for column generation and topological ordering in crew/tail pairing in Java• Designed and tested an algorithm for creating monthly schedules of crews and aircrafts given constraints such as shift times, time spent at each location, qualifications and work rules such as seniority• Tested and optimized linear programming models using IBM ILOG CPLEX Optimization Studio