Mohsen Dehghankar
Computer Science PhD Candidate | Chicago, IL
About
I am a 4th-year Ph.D. candidate in Computer Science at the University of Illinois Chicago, advised by Prof. Abolfazl Asudeh. I completed my B.Sc. in Computer Engineering, with a minor in Mathematics, at Sharif University of Technology.
My research focuses on algorithm design for improving different stages of the machine learning and data pipeline. Broadly, my research spans two directions:
LLM Inference Efficiency: Designing algorithms that accelerate the inference in LLMs.
Retrieval Problems: Problems such as Approximate Nearest Neighbor (ANN) search and ranked retrieval. Exploring how retrieval can be used to enhance inference.
In addition, I have an interest in computational geometry algorithms, particularly in their applications to data science and database problems.
Thesis (Proposal) Title: “From Data to Models: Efficient Algorithms for Retrieval and Inference”
Recent News
- Joined Adobe Research as a Research Scientist Intern, Summer 2026.
- RSR-core is accepted to VLDB 2026 (Demo Track)!
- Random-Access Ranked Retrieval is accepted to KDD 2026!
- Received the Graduate Student Award for Exceptional Research Promise, UIC College of Engineering.
- Gave a talk at the Midwest DataDay seminar, University of Chicago, on “Hierarchical Epsilon-Net Graphs for ANN”. [Slides]
- Defended my thesis proposal and advanced to Ph.D. candidacy.
Selected Papers
Sparse Attention as a Range Searching Problem: Towards an Inference-Efficient Index for KV Cache
Mohsen Dehghankar, Abolfazl Asudeh
Under Review (2026) LLM EfficiencyRetrieval
[PDF] | [Code]Random-Access Ranked Retrieval and Similarity Search
Mohsen Dehghankar, Abolfazl Asudeh, Rahul Mittal, Suraj Shetiya, Gautam Das
KDD 2026 Retrieval
[PDF] | [Code]RSR-core: A High-Performance Engine for Low-Bit Matrix-Vector Multiplication
Mohsen Dehghankar, Abolfazl Asudeh
VLDB 2026 (Demonstration Track) LLM Efficiency
[PDF] | [Code]An Efficient Matrix Multiplication Algorithm for Accelerating Inference in Binary and Ternary Neural Networks
Mohsen Dehghankar, Mahdi Erfanian, Abolfazl Asudeh
ICML 2025 LLM Efficiency
[PDF] | [Code] | [Project Page]Hierarchical Epsilon-Net Graphs: Time Guarantees for HNSW in Approximate Nearest Neighbor Search
Mohsen Dehghankar, Abolfazl Asudeh
Preprint (2025) Retrieval
[PDF] | [Code] | [Slides]An Adversary-Resistant Multi-Agent LLM System via Credibility Scoring
Sana Ebrahimi, Mohsen Dehghankar, Abolfazl Asudeh
AACL 2025
[PDF]Rank It, Then Ask It: Input Reranking for Maximizing the Performance of LLMs on Symmetric Tasks
Mohsen Dehghankar, Abolfazl Asudeh
KDD 2025 Retrieval
[PDF] | [Code]Mining the Minoria: Unknown, Under-represented, and Under-performing Minority Groups
Mohsen Dehghankar, Abolfazl Asudeh
VLDB 2025
[PDF] | [Code]Fair Set Cover
Mohsen Dehghankar, Rahul Raychaudhury, Stavros Sintos, Abolfazl Asudeh
KDD 2025
[PDF] | [Code] | [Video]
Contact
You can reach me by email at mdehgh2@uic.edu. You can also connect with me on LinkedIn, GitHub, and X.