Stefan Gramatovici
Stefan Gramatovici is a data scientist working on ChatGPT performance as a Member of Technical Staff at OpenAI (Greater Seattle Area). He studied Mathematics at Harvard University (B.A., 2014-2018) and previously worked as a Data Scientist at Google (Chrome Data Science, experimentation and statistical modelling, 2022-2026) and at Microsoft (2018-2022), including the Windows Experimentation Platform and the IDEAs team. In 2016 he interned as a Quantitative Research Intern at Ellington Management Group on automated trading strategies for equities and structured credit products.
- Quant employment verification: CONFIRMED. LinkedIn profile lists a 'Quantitative Research Intern' role at Ellington Management Group (2016) working on automated trading strategies for equities and structured credit products; Ellington Management Group is a registered investment adviser / hedge fund founded 1994 per its own site, so the employer genuinely operates in quantitative finance. This is an internship (May-Aug 2016 per caller lead), not full-time, and it rests on index/LinkedIn data rather than a separate primary-source employment record.Jun 1, 2016
- Identity anchor: official IMO record confirms Ştefan-Cosmin Gramatovici represented Romania at IMO 2013 with a bronze medal (rank 211, 17 points).Jul 1, 2013
- LinkedIn/index profile shows 'Stefan Gramatovici', Harvard Mathematics class of 2018, now Member of Technical Staff at OpenAI in Greater Seattle; the Romanian name plus Harvard-math timeline is consistent with the IMO 2013 Romanian contestant, but no single source yet directly joins the two identities.Sep 1, 2026
- Career timeline (index/LinkedIn-sourced): OpenAI Member of Technical Staff (2026-), Google Data Scientist Research (2022-2026), Microsoft Data Scientist II (2020-2022), Microsoft Data Scientist (2018-2020), Microsoft Data Science Intern (2017), Ellington quant research intern (2016).Sep 1, 2026
- Co-authored 'Novelty and Primacy: A Long-Term Estimator for Online Experiments' (arXiv 2102.12893) on long-term effects in online experiments while at Microsoft.Feb 1, 2021
Data Science for ChatGPT performance.
Working in Chrome Data Science on experimentation and advanced statistical modelling.
IDEAs team: Built a hierarchical forecasting model for business time series such as number of M365 Subscribers. Advised on experiment best practices. CosmosDB: Developed a new heur
Worked on WExp, the Windows Experimentation Platform. Built model to identify and estimate long term effects in experiments. See https://arxiv.org/abs/2102.12893
Build a prototype differential privacy library. It allows querying of customer data with managed privacy loss under the server-side differential privacy model.
Quantitative Research Intern working on automated trading strategies for equities and structured credit products
Competition record
Romania · IMO