Ozan Ogreden Email & Phone Number

Utrecht, Netherlands

My main responsibility was leading cross-discipline projects with the finance team: SnappCar's Finance System is a key asset for the business. It's a conceptually beautiful, bespoke software system for multi-party transactions.Challenges:Legacy codebase. Low resolution understanding of the system. Weak communication lines between finance and product teams. Discrepancy between actual financial workflows and how the system behaved. Unclear/unmeasured project goals in this domain. Key results:- Enabled use of customer deposits for financial workflows. Leading to € XX.000 / month use of deposits instead of "after-charges" (= after the initial payment for a rental). Order of magnitude decrease in debt per rental. Order of magnitude decrease in debt per aftercharge.- Simplified internal financial interfaces to accomodate current workflows. Operational teams needed to use a complex "swiss knife" to serve our customers. Complex tools increased the risk for mistakes, which increased the risk of customer payouts being delayed. I identified problems, prioritised and implemented simplifications. (Impact being measured.)Currently:- Extending system to accomodate current business needs.- Exploring customer facing product development opportunities with MVPsWhile delivering these results, I also:- Improved our unit testing approach by introducing modern tooling when legacy, in-house tools were not serving us.- Advocated and introduced the use of design patterns to address developer pain-points that resulted from organic development

Amsterdam, North Holland, Netherlands

Amsterdam Area, Netherlands

Amsterdam Area, Netherlands

I was the first technical hire of the Marketing Technology team. I took on a wide variety of roles in this role. As a data scientist:- Developed high level wrappers for Google Ads API around the official Python SDK, implemented a number of minor projects. Estimated time saving is ~3 hours/week for the total a team of <5 marketers.- Lead two major projects from specification to implementation:1. A tool that detects high performance search terms, ensures their uniqueness against thousands of keywords in our accounts, creates keywords and ads. Built using our Google Ads API wrapper, `scipy` stack and open-source NLP libraries. Results: This task has added more than 7.5k spell checked unique new keywords to Albelli accounts and created about 20k ads. Saving of ~2k per week by enabling actions by a 3rd party bid optimization tool.2. A tool that optimizes the spend for strategically important parts of our accounts, using Bayesian GLMs. Built using Stan, `brms` and our Google Ads wrappers. It's still under evaluation: overall satisfied with the decisions from a marketing perspective but room for improvement for realizing saving potential.Next to these, I set up a version controlled, containerized development and production workflow with the IT team (using git, Docker, Jenkins, AWS ECS and Terraform); I was involved with hiring and onboarding of a second data scientist; supported analysts and consultants.

Amsterdam, Netherlands

I worked as a junior researcher as part of the Dutch-Flemish PROMIS groups. After 12 months, I decided to quit the academic trajectory. During these 12 months:- I developed dcurver - an R package that provides C++ utilities for the Davidian curves (Zhang & Davidian, 2001). A Davidian curve defines a seminonparametric density, whose shape and flexibility can be tuned by easy to estimate parameters. The package provides the density function, the gradient of the loglikelihood and a random generator for Davidian curves.- I contributed dcurver to R package mirt (at version 1.28).- I developed and internally presented several lectures for a graduate level course on Item Response Theory.- I initiated a standardization of the R code base of the Dutch-Flemish PROMIS group.- I performed research and research support activities within the Epidemiology & Biostatistics Department.

Providing human evaluation input to search result evaluation algorithm

Building predictive models for stress using real-life sensor data.

İstanbul, Türkiye

Utrecht Area, Netherlands

Investigation of the relationship between white matter microstructure and cognitive performance in a longitudinal data set

Master Of Science (M.Sc.), Methodology And Statistics For Behavioral, Biomedical And Psychological Sciences, 8.2/10.0

Master's thesis: "Variable Selection Strategies in Multinomial Prediction Models"

Bachelor Of Arts (B.A.), Psychology, 3.82/4.00, 3Rd Of Class

Activities and Societies: BU Cognitive Processes Lab, Research Assistant (2013 - 2015)Poster presentation: Yildirim, I., Öğreden, O. &.

Education record