Keivan Rezaei

Keivan Rezaei is a PhD candidate in Computer Science at the University of Maryland, College Park (Reliable AI Lab), supervised by Prof. Soheil Feizi and Prof. Mohammad Hajiaghayi. He works on interpretability of generative AI, machine unlearning, knowledge localization, model editing, and economics+AI. He earned a BS in Computer Engineering at Sharif University of Technology (rank 1st, 2018-2022) and an MS in CS at UMD (2022-2024). A competitive-programming background includes a silver medal for Iran at the 30th International Olympiad in Informatics (IOI 2018) and ICPC World Finals appearances. Research internships have spanned Google Research, Adobe, Allen Institute for AI, EPFL, and - in Summer 2026 - an ML research internship at Susquehanna International Group (SIG), a quantitative trading firm.

Notes7

• Quant employment verification: CONFIRMED. CV and personal site document an actual ML Research Internship at Susquehanna International Group (SIG), Jun 2026 - Aug 2026, optimizing data selection for intra-day price prediction models; SIG is a proprietary quantitative trading firm, per SIG's own careers listing describing it as 'a global quantitative trading firm'. Identity is supported by mutual cross-links between his CV, personal site, LinkedIn and UMD profile.Aug 2026
• The SIG ML internship is quantitatively applied: the CV states the work optimized data selection for intra-day price prediction models, i.e. ML for trading signals rather than a generic software internship.Jun 2026
• The 2026 SIG start date on his own site ('I'll be joining Susquehanna International Group for Summer 2026') and completed CV entry precede the current date of Sep 2026, so the internship is best read as completed, not ongoing.Jun 26, 2026
• Past employers are research/industry-lab roles (Google Research Student Researcher Feb-Apr 2025; Adobe Research Intern May-Aug 2025; Ai2 Research Intern May-Oct 2024; EPFL ML theory intern Jul-Sep 2021) rather than additional finance roles, so SIG is the only quantitative-finance employer on record.
• Research at UMD sits in the Reliable AI Lab supervised by Soheil Feizi and Mohammad Hajiaghayi, spanning interpretability, unlearning and economics+AI - a breadth (ML + algorithmic game theory) that predates any finance work.
• Ranked 1st in his BS cohort (Computer Engineering, Sharif University of Technology, 2018-2022) before starting a CS PhD at UMD in 2022; he completed an MS at UMD in 2024 en route.Sep 2022
• Represents Iran at the top of competitive programming: silver medal at IOI 2018 (30th IOI) as an official contestant, plus APIO 2018 silver; he later served as Iran's team leader at IOI 2021.Sep 2018
• First-author work on machine-unlearning evaluation: 'RESTOR: Knowledge Recovery in Machine Unlearning' and 'Revisiting the Past: Data Unlearning with Model State History' (ICLR 2026, co-first author with Mehrdad Saberi).
• Public contact: [contact omitted] (UMD directory) and phone +1 (240) 413-8060 as self-published on his CV, with office IRB 4120.
• His X bio ('CS Ph.D. Student @UofMaryland || Prv. ML Research Intern @ SIG || Prv. Student Researcher @GoogleAI || Prv. intern @allen_ai Ai2 || Alum @SharifSocial') is a self-reported identity cross-check that independently matches the CV employers and Sharif background.
• Strong ICPC record: 2023 World Finalist (could not attend due to visa issues), 3rd at the 2023 ICPC North America Championship, 33rd at the 2020 World Finals, and 1st at the 2019 Asian Regional Contest.Jan 2023
• Publishes at top ML venues (ICML 2023, ICLR 2024, ICML 2024, NeurIPS 2024, NeurIPS 2025, ICLR 2026) across interpretability, knowledge localization and LLM ad auctions.