Mushtari Sadia
About me
I'm Sadia (she/her), a second-year PhD student at the department of CSE in University of Michigan, Ann Arbor, working as a Research Assistant @ Michigan Systems Laboratory, and I'm co-advised by Professor Ang Chen and Professor Amrita Roy Chowdhury.
My research interest lies in reimagining core systems problems using computer vision, natural language processing, and multimodal models, to build secure and/or more capable systems. Therefore, I am especially motivated by problems at the intersection of (1) systems/security, and (2) AI.
I previously graduated from the department of CSE, BUET, and worked as a lecturer in the department of CSE, BRAC University, and as an adjunct lecturer at the department of CSE, BUET.
In my free time, I love travelling, dancing and binge-watching sitcoms. 
Work Experience
-
Graduate Student Research Assistant
Advisors: Ang Chen, Amrita Roy Chowdhury.
Department of Computer Science and Engineering,
University of Michigan, Ann Arbor
August 2024 — Present -
Lecturer
Department of Computer Science and Engineering,
BRAC University
June 2023 — July 2024 -
Adjunct Lecturer
Department of Computer Science and Engineering,
Bangladesh University of Engineering and Technology
November 2023 — March 2024
Research Experience
(* denotes equal contribution)
-
SQUiD: Synthesizing Relational Databases from Unstructured Text (2025)
Mushtari Sadia, Zhenning Yang, Yunming Xiao, Ang Chen, Amrita Roy Chowdhury
EMNLP 2025 (Oral Presentation, 2.75% selection rate). [Paper]Designed a novel neurosymbolic framework that automatically synthesizes relational databases from unstructured text by generating the schema and populating its tables from raw text. Designed a four-stage pipeline, achieving up to 91% entity coverage with 100% primary and foreign key validity, and improving up to 18× over direct LLM prompting baselines.
-
Pimϵnto: A Privacy Framework for Querying Text (2026)
Mushtari Sadia, Ang Chen, Amrita Roy Chowdhury
Under review at USENIX Security 2027.Developed Pimϵnto, which grounds natural language text, queries, and privacy policies into a relational database to integrate, for the first time, contextual integrity and differential privacy over unstructured text. Introduced contextual differential privacy to reduce noise and a privacy-aware Text-to-SQL method using Monte Carlo Tree Search that selects the lowest-noise query in 75.3% of cases while achieving 0% leakage.
-
SPARC: A Multi-Agent System for Electrical Circuit Question Answering (2026)
Mushtari Sadia, Zhenning Yang, Umme Habiba Lamia, Nishat Shawrin, Ang Chen, Amrita Roy Chowdhury
Under review at AAAI 2027.Designed a multi-agent LLM system with orchestration that answers questions on electrical circuit diagrams by synthesizing and executing SPICE simulations. The system improves the reliability and interpretability of multimodal LLM reasoning on this task, reaching 83% accuracy with up to 58% improvement over baselines.
-
Trinqϵt: Private Triangle and Quadrangle Counting over Distributed Graphs (2026)
Mushtari Sadia, Amrita Roy Chowdhury
Under review at NDSS 2027.Developed Trinqϵt, a privacy-preserving graph analytics system using the MP-SPDZ library, for triangle and quadrangle detection over distributed graphs with malicious security. Introduced MPC algorithms that exploit graph sparsity through differentially private noisy degree release, achieving 105× improvements in runtime.
-
Nomai: Single-Server Federated Learning with Private Weights (2026)
Yiyi Sun, Mushtari Sadia, Amrita Roy Chowdhury
Under review at USENIX Security 2027.Developed Nomai, a scalable MPC-based federated learning system that provides data privacy, model privacy, and integrity for single-server deployments with up to 106 users, and can train MNIST in under 50ms.
-
LogShield: A Transformer-based APT Detection System Leveraging Self-attention (2022)
Sihat Afnan*, Mushtari Sadia*, Shahrear Iqbal, Anindya Iqbal
[Preprint]Undergraduate thesis project under Dr. Anindya Iqbal and Dr. Shahrear Iqbal (Research Officer, National Research Council (NRC) Canada). I co-implemented a framework that creates provenance graphs from raw DARPA OPTC and DARPA TC E3 log data, generates event sequences from those graphs, and transforms them for transformer-based language models such as BERT, RoBERTa, and GPT. My contributions included building the PostgreSQL database from the raw logs, designing SQL queries to generate the provenance graph, co-writing the graph preprocessing code, and building the experiments with various pre-trained LLMs. The framework achieved state-of-the-art performance in detecting APT attacks.
-
Multi-modal Swarm Intelligence for Secure UAV Missions (2024)
Yunming Xiao, Mushtari Sadia, Ang Chen
GenZero Workshop 2024 (Poster)Unmanned Aerial Vehicles (UAVs) have found wide use in various tasks, but developing swarm intelligence for intrusion detection is far from easy. In this project, we proposed to leverage recent advances in large multimodal models (LMMs) that can fuse multiple data sources for secure missions. Our project will fine-tune and deploy LLMs of varying sizes to the edge/fog UAVs, combining data sources such as sensory inputs (e.g., camera, IMU) as well as internal operational data (e.g., syscall logs). This will enable a real-time detection and response system to thwart threats with swarm-wide coordination.
-
Advancing Parallelization in Deep Learning Training: A Novel HSIC-Based Approach and Its Comparative Performance Against Traditional Federated Models (2023)
Mushtari Sadia, Praneeth Vepakomma
Worked with Dr. Praneeth Vepakomma (MIT Camera Culture Group) as part of the Fatima Fellowship Research Program, on a distributed machine learning project. We developed a method for parallelizing the forward propagation in neural network training, utilizing the HSIC objective to eliminate the need for backpropagation, while maintaining the same level of accuracy. We also found that, in some instances, employing slightly outdated local updates can significantly reduce communication costs without compromising accuracy.
Education
-
PhD in Computer Science and Engineering
University of Michigan, Ann Arbor
Advisors: Ang Chen, Amrita Roy Chowdhury.
August 2024 - Present -
M.Sc. in Computer Science and Engineering
University of Michigan, Ann Arbor
2026 -
B.Sc. in Computer Science and Engineering
Bangladesh University of Engineering and Technology
April 2018 - May 2023
Technical Skills
-
Programming Languages
C/C++, x86 Assembly, Bison/Flex, Python, Java, Javascript, Bash, MySQL
-
Frameworks
Docker, PyTorch, NS3, xv6, Django REST, ReactJS, MP-SPDZ, Git, Oracle DBMS, LaTeX
-
Libraries
Sklearn, Pandas, Matplotlib, Seaborn
