Questions worth
investigating.

Institutional PE research with an external industry partner: extracting, validating, and standardizing confidential deal and valuation data, and time-series modeling across portfolio companies to improve methodological rigor and data transparency.

A research design for bringing generative AI into behavioral economics: using generative models to create synthetic consumer profiles, then exposing them to behavioral nudges in a simulated environment to test which interventions shift decisions before anything reaches real people. The work pairs this with a serious ethical framework (informed consent, anonymization, algorithmic-bias mitigation) and a hybrid validation strategy, synthetic data for fast, low-risk hypothesis testing followed by traditional lab and field methods to confirm the results hold in the real world.

A controlled experiment run months after ChatGPT's release: students wrote the same history assignment twice, with and without the tool, and language instructors graded both versions blind to which was which. The result cut against the hype, and prefigures a theme I would return to in my thesis: the AI-assisted papers averaged no higher (48% versus 51%), because a fluent, generic voice graded worse than an authentic one. Coded interviews with the participants mapped where the tool genuinely helped (summarizing, brainstorming, time management) against where it failed (reliability, fabricated detail, academic-integrity concerns), an interdisciplinary study across computer-science and social-science students.

Part of a study asking whether entrepreneurial traits can be trained. A cross-country literature review, deliberately spanning developed and developing economies to avoid geographic bias, converged on grit and perseverance as the traits most predictive of entrepreneurial success. The team then built a custom game to measure it: a deliberately frustrating, Flappy-Bird-inspired catch-the-ball prototype in Python, where how long a player keeps chasing a randomly-moving target stands in for perseverance under uncertainty, ready to be fielded pre- and post-training to test whether the trait shifts.

What venture capitalists look for in startups, the trade-offs of raising VC funding, and a documented map of the Turkish venture-capital ecosystem and its common exit strategies.

A solo empirical-macroeconomics project on the marginal propensity to consume in Turkey, the share of an income change that is spent rather than saved, which sits at the heart of the fiscal multiplier and of how tax and spending policy move aggregate demand. The work gathers the literature on MPC, identifies a current framework for estimating it in a developing-country setting, and reweighs earlier findings against new empirical evidence.

A mixed-methods study of how the sustainability disclosures of major international fashion brands actually affect their suppliers. Starting from the Fashion Transparency Index 2020, the work narrowed 250 brands to the Turkish suppliers serving several of them at once, combined that quantitative disclosure data with in-depth supplier interviews, and surfaced a sampling paradox worth its own study: because disclosed suppliers are pre-selected for good practice through brand codes of conduct, they cannot reveal the effect on the firms hiding poor practice. The recurring theme across cases was that audits, not disclosure, are the real lever, since their absence lets non-compliant competitors undercut the firms paying to be sustainable.

A testing approach that lets non-technical stakeholders test their own mobile apps without a developer, by pairing the visual, block-based simplicity of Google Blockly with the power of professional tooling. The team authored behavior-driven (Given / When / Then) test scenarios by recording capture-and-replay scripts in TestProject and running them through a modified Blockly environment against Android apps on an emulator. Timing data confirmed the core problem, that users unfamiliar with code were markedly slower with professional tools, and pointed to a visual interface as a plausible way to close the gap.

Original geometry and number-theory research on regular, semi-regular and isovalent tilings, developed over two years as a mathematics-olympiad student. The work uses Euler’s formula, Diophantine equations, and GeoGebra to classify and construct polygon configurations. Awards include an ICYS bronze medal, TÜBİTAK regional first place, and MATDER gold.