research projects

Civic Health and Insitutions Project (CHIP50)

This project is a big, NSF-funded collaborative effort of Harvard, Northeastern, Northwestern, and Rutgers. We examine social behaviors, social networks, political preferences, and the impact of messaging and regulation on individual/community outcomes through large-scale online surveys. We share findings directly with the public, the media, and federal/state officials. I have several responsibilities.

I supervise data collection from 25,000 Americans with Census-based demographic quotas in 50 states plus the DC each wave, and collected 600,000 responses from 2020 to present (2024) through 31 survey waves so far. I am also responsible for managing our budget, the compensation we pay to each respondent, by assessing the recruitment numbers each day and their distribution over time and states. Stabilizing the recruitment traffic is important to get quality answers and reduced bias, especially when you recruit tens of thousands of respondents approximately every 2 months.

When data collection is done, I use python, R and statistical methods to analyze survey data, often combine them with other datasets, such as social media accounts we collect from people or other third-party data sources (e.g., waste water surveillance, voter data, deaths and hospitalizations in states etc.). I come up with insights and plots.

Then, I write [Google Scholar] academic papers, reports and op-eds, or help others write as their co-author. Our works are covered by outlets such as the Atlantic, NYT, Politico, WP, WSJ, and published by JAMA, PNAS, Nature, Science etc.

Amazon review networks as a recommendation system

I created a network-based recommendation system using Amazon product reviews (data from Jianmo Ni et al.). I hypothesized that there exists preference transitivity among Amazon customers through homophily and triadic closures and that these can be used as signals to predict future purchasing behavior and preferences. I verified my hypothesis using graph theory and network algorithms in certain product categories.

Automatic Attendance

I developed a facial recognition app with Dr. Meltem Tolunay, which takes real-time attendance from classroom photos and emails to lecturer. We used one-shot learning and Convolutional Neural Networks to leverage the student photos already stored in university databases for transfer learning. Used Python, TensorFlow, OpenCV, Django.

Take us back to times we take classes together. Please.

SkipSim: Scalable Skip Graph Simulator

I implemented (Java) tests for ‘SkipSim’, a simulator that enables skip-graph-based algorithms (e.g., blockchains, P2P cloud storage) to be simulated while preserving their scalability and decentralized nature.

Circular permutations of identical objects

I developed a novel mathematical approach to enumerate circular permutations of identical objects accounting for their reflections as well (‘bracelets’ or turnover 'necklaces'), and introduced new formulas in Combinatorics. My formulas can be used to count certain isomorphisms, such as alkane (paraffin) numbers. I received the Karl Menger Award from American Mathematical Society at the Intel Science and Engineering Fair in Los Angeles. If interested, you can access them on OEIS with serials: A210464, A005995, A005994, A018210. Some news on national newspapers and tech magazines are here: 1, 2, 3, 4, 5, 6

Simple example: Number of different bracelets with 3 red and 3 green *identical* beads is 3. This is because we allow necklaces to be turned over: we count the reflections as the same bracelet. If we didn't allow turnovers, there would be 4 different 'necklaces' (4th one being the 2nd one turned over)