Selim Yaman
Data Scientist
I’m a Senior Data Scientist at Office for National Statistics, and I deal with financial data.
Previously, I was a postdoc at the University of Mannheim’s Center for Social Data Science, working with Dr. Marc Ratkovic to see how far we can push LLMs in the social sciences. I earned my PhD in Political Methodology from American University, where I spent a lot of time thinking about data (and perhaps a little too much time with Bayesian ways of modeling data) as well as previous, not-so-large language models. Before that, I produced political satire TV shows in London, UK.
Experience
Post-Doctoral Researcher
- Conducting advanced research on large language models (LLMs) in collaboration with Dr. Marc Ratkovic, with a focus on applications in political methodology. Teaching data science courses.
Replication Analyst
- Worked in the most reputable academic journal in political methodology
- Evaluated statistical analyses of submitted manuscripts, and assisted authors in debugging code
- Heavily used command line tools, remote servers, SQL, R, and Python to ensure that all code reproduce authors’ original results
Graduate Assistant
- Organized and thought at data science courses at Institute for Data Science and Big Data
- Teaching Assistant for advanced Bayesian Statistics & Introduction to Statistics course
- Helped students in understanding machine learning, coding, statistical modeling, debugging; held regular office hours; graded code assignments
- R package development and maintenance (BaM, and ProtectR)
Data Analyst and TV Producer
I analyzed and visualized data on the company’s international auidence. I also helped producing shows.
Education
PhD in Political Science
- Subfields: Methodology and Comparative Politics
- Dissertation on NLP methods in Political Science
- Bayesian Statistics
- Machine Learning
- Advanced Statistical Analysis
- NLP
MSc in Political Economy of Development
- Dissertation on Religion Effect on Economic Growth
- Microeconomics
- STATA, R
BA in Economics
This is where I first learnt about the introduction of data science with plenty of econometrics classes. Bogazici University is the most prestigious school in Turkey.
Accomplishments
TA at ICPSR
I was a TA for the machine learning course at University of Michigan’s ICPSR, worked with Professor Christopher Hare
Conference Presentation
I presented a research article titled Bridging the Theory and Practice: Introducing Bayesian Partially-Protected Lasso’ at PolMeth 2023, Stanford University
Participant at ICPSR
Intensive summer program for methodology courses. I took two classes here: Machine Learning, and Python.
Teaching Assistant at Institute for Data Science and Big Data
In this every-day intensive course, I helped students everyday in debudding code, finding relavant datasets, and doing statistical analysis.
Teaching Assistant at Bayesian Statistics
I held regular office hours, designed and evaluated Bayesian statistics assignments every week. We used JAGS to implement multilevel Bayesian modeling.
Fellowship Award
I won the prestigious Kerwin Doctoral Fellowship - 10,000 USD for PhD dissertation
Completed Summer Institute in Computational Social Sciences (SICSS) Istanbul
I participated in the summer courses, and we wrote a short research project on tweets posted on Israel-Palestine conflict. We estimated ideologies of Twitter users, and extracted what topics that they discussed.
Conference Presentation
I presented a research article titled ‘Estimating User Ideology on Twitter Using Large Language Models’ at the Comptext 2023 in Glasgow, UK
Conference Presentation
I presented a research article titled ‘Political Economy of Warlords: A Local-Level Analysis from Afghanistan’ at the 78th Annual Midwest Political Science Conference. You can find the article here.
Completed Winter Institute in Computational Social Sciences (WICSS) Tuscon
I participated an online program on computational social sciences hosted by University of Arizona.
Teaching Assistant at Introduction to Statistics using R
I held weekly office hours to help students overcome programming and statistical modeling challanges.