• Given an egocentric video as input, the task is to predict the action category of each frame, the start and the end boundaries of all segments that include an action instance, and the action category of each segment that is predicted as an action instance.

• Prediction and interpretation of the effect of copy number aberration on gene expression of tumor samples.• Extended my studies to do an extra research project.

• By multiplying the gradients from a Score-based Generative Model with the Jacobian of a 3D rendering function, filled the missing regions of a given UV face texture. The objective is to render the face from arbitrary viewpoints.

• Worked on implementation of a product that lets users measure their blood pressure by putting their fingertip on a smartphone camera.• Used Python, PyTorch, Amazon Web Services.

• Predicted canonization scores for historical English and German books using textual information.• Paper was accepted for publication in Computational Humanities Research 2021 and Conference of Computational Literary Studies 2022.

• Worked on implementation of a product which lets users measure their vital signs by putting their fingertip on a smartphone camera or recording a face video.• Implemented the back-end of a real-time vital sign detection product using Flask, Socket.io, Amazon Kinesis Video Streams, AWS Lambda, AWS Fargate.• Deployed research software prototypes in AWS. Translated these prototypes from Python to Java so that the mobile developers can deploy them on mobile devices. Supervised interns for both tasks.• Used the following AWS services: Lambda, DynamoDB, Fargate, Amazon Kinesis Video Streams, S3, EC2, ElastiCache, RDS, VPC, SageMaker, CloudFormation, CodePipeline, CloudFront.• Decreased monthly AWS costs by 31%. Decreased error rate in heart rate detection algorithm by 75%.• Worked full-time for 7 months, part-time for 3 months.

• Implemented a time series forecasting library to predict the next day’s amount of cash deposit and withdrawal in Turkey branches of a global bank. Used PyTorch, Python, R. This project increased their gain on overnight interest income by 30%.• Implemented an NLP library for the Turkey branch of a global e-commerce company which improved their auto-correct feature and search results. Used Java, Python and Gensim. Increased the usage of ’Did you mean’ feature on their website by more than 200%.• Used AWS, MySQL, Redis, Node.JS and Serverless Framework to deploy machine learning models into production environment.• Worked full-time for 13 months, part-time for 9 months.

• Worked on the implementation of a platform that lets users analyze and visualize large textual data.• Used Python, PySpark, Node.JS, UnitTest, Docker.• Worked full-time for 2 months (1 month in Stockholm, Sweden), part-time for 8 months

• Implemented a natural language processing platform that lets users create their own chatbots. Used Java and MongoDB.• Implemented a Facebook Messenger chatbot that lets users query the location of the nearest ING branch. Used Node.js and Heroku.