Advanced Search with Generative AIPeruser AI scours the internet, reads and summarizes relevant webpages, and gathers a coherent answer with links, references and images for your search queries.

Led the Qualified Applicant AI model which provide recommendations to hirers and job-seekers. I operate end-to-end from ideation and modeling to implementation and productionization.• Built an end-to-end GBDT model, productionized and launched in two novel applications: highlighting promising jobs to members; showcasing matching candidates to recruiters. Led cross-team efforts to ramp in production. Achieved significant gains in user engagement and job marketplace efficiency.• Designed TensorFlow / Deep Learning architecture for the Qualified Applicant model, its surrounding tech stack, and its launch in four product verticals.Focus areas: Recommender Systems, Wide & Deep Learning, Generalized Linear Mixed Models, Gradient Boosted TreesTech Stack: Apache Spark, Hadoop, Scala, Python, Java EE, Gradle, Groovy, TensorFlow, GLMix, XGBoost

Totient develops computation-based experimentation platforms to facilitate early stages of the drug discovery process. I contributed to R&D efforts by applying machine learning and statistics techniques to large cancer genomics and pharmacogenomics (PGx) datasets.• Built a machine learning model and its data pipeline which predicts gene-drug interactions from large PGx datasets. It produced drug target candidates for investigation of biologists (out of 20.000+ protein-coding genes). Used Elastic Net and ANOVA.•Wrote a white paper showcasing advantages of the company's drug discovery strategy against the SOTA methods (based on my analysis on the gene-drug datasets, and review of the industrial methods).• Developed and maintained company-wide data standardization and preprocessing software for cancer genomics datasets. It saved time for colleagues through automatization.• Designed algorithms to identify cancer driver genes based on the clustering of mutations. Achieved highest precision among similar-purposed methods.Focus Areas: linear regression, elastic net, data pipeline design & automation on cloud, data standardization, ANOVA, MCMC algorithms, Dirichlet-multinomial regression, correlation analysesTech stack: Python, R, C++, scikit-learn, pandas, Docker, AWS

Developed an AutoML tool during my 3-weeks fellowship. It discovers dataset-specific image augmentation strategies for CNN’s by utilizing Bayesian Optimization. The tool: • Reduced error by 60% for the Wide-ResNet model on a CIFAR10 reference image dataset. • Achieved >50x speed increase compared to previous solution–AutoAugment from Google.https://github.com/barisozmen/deepaugmentFocus areas: AutoML, Bayesian Optimization, Convolutional Neural Networks, Wide Residual Networks

Researched the navigation system of the brain. Developed a statistical method for determining firing-field shapes of the grid cells within the entorhinal cortex. My work included in https://abstracts.g-node.org/abstracts/607e3ad9-51a0-4c51-809c-ef93ff7ec551

Researched clustering algorithms (particularly k-means, DBSCAN, EM, and hierarchical clusterings) to be integrated in the company’s main ML model which predicts real-estate prices in Istanbul. Tech Stack: C++, LIBSVM