I'm a Computer Science Ph.D. student at the University of Maryland, advised by Ramani Duraiswami. My research focuses on Transformer efficiency (sub-quadratic attention, KV-cache reduction), post-training optimization of Transformers, and prediction models.
Experience
-
Summit Securities Group, HFT ML Research Intern2026Short-horizon price prediction from high-frequency market data with gradient boosting and Transformers.
-
Apple, AI/ML Engineer Intern2026Agentic multimodal RAG system for the RF Systems team.
-
University of Maryland, Graduate Research Assistant2023–Efficient Transformers and their applications.
-
University of Maryland, Software Developer2022–23
Energy monitoring tool for State of Maryland buildings.
Selected publications
-
Iceberg Attention: Sublinear Long-Context Decoding with Mixture-of-Gaussians
A. Gerami, L. Koroshinadze, S. Madani, P. Pandey, R. Duraiswami
Preprint, 2026
-
On the Application of Linear Attention in Multimodal Transformers
A. Gerami, S. Madani, R. Duraiswami
CVPR 2026
-
Auditing Algorithmic Bias in Transformer-Based Trading
A. Gerami, R. Duraiswami
NeurIPS 2025 Workshop on Generative AI in Finance
-
Two-Level Generative Evolutionary Prompt Adaptation for Non-Parametric Continual Learning
A. Gerami et al.
Preprint
-
Efficient Spatial Audio Rendering via Differentiable FIR-to-IIR Estimation
A. Gerami, B. Zhi, D. N. Zotkin, R. Duraiswami
ICASSP 2025
-
GUST: Graph Edge-Coloring Utilization for Accelerating Sparse Matrix-Vector Multiplication
A. Gerami, B. Asgari
ASPLOS 2024
Education
- Ph.D., Computer Science, University of Maryland2023–
- M.Sc., Electrical Engineering, University of Maryland2022–23
- B.Sc., Electrical Engineering, Sharif University of Technology2016–20
Honors
- NSF NeuroPAC Fellowship2025
- Outstanding Graduate Research Assistant Award, UMD2024
- Ranked 21st among ~250,000 in Iran's national university entrance exam2016