University of California, Santa Barbara

Portrait of Sajjad Ghiasvand

Sajjad Ghiasvand

PhD Student, Electrical & Computer Engineering

University of California, Santa Barbara

About Me

I am a PhD student in the Electrical and Computer Engineering Department at UC Santa Barbara, advised by Prof. Ramtin Pedarsani and Prof. Mahnoosh Alizadeh. Prior to joining UCSB, I earned my B.Sc. in Electrical Engineering with a minor in Computer Science from Sharif University of Technology in 2023.

My research spans text-based and multimodal large language models, with emphasis on:

  • Personalization of LLMs and VLMs
  • Efficient fine-tuning of LLMs and VLMs
  • LLM post-training & imitation learning
  • Federated and decentralized learning
  • Robust machine learning

Feel free to reach out if you are interested in discussing a collaboration!

News

  • Jun 2026 "MMLoP: Multi-Modal Low-Rank Prompting for Efficient Vision-Language Adaptation" accepted to ECCV 2026.
  • Jun 2026 Started a Research Internship at Higharc, working on vision-language models for agentic home design editing.
  • Mar 2026 "Few-Shot Adversarial Low-Rank Fine-Tuning of Vision-Language Models" accepted to Trustworthy AI Workshop at ICLR 2026.
  • Jan 2026 "pFedMMA: Personalized Federated Fine-Tuning with Multi-Modal Adapter for Vision-Language Models" accepted to ICLR 2026.
  • Aug 2025 Started an ML Internship at Handshake AI (joint project with OpenAI).
  • May 2025 "Communication-Efficient and Tensorized Federated Fine-Tuning of Large Language Models" accepted to ACL Findings 2025.
  • May 2025 "Decentralized Low-Rank Fine-Tuning of Large Language Models" accepted to REALM Workshop at ACL 2025.
  • Apr 2025 "Robust Decentralized Learning with Local Updates and Gradient Tracking" accepted to IEEE Transactions on Networking.

Experience

  • Research Intern — Higharc

    Jun 2026 – Present

    Built data pipelines for structured architectural datasets; adapted vision-language models for agentic home design editing. Implemented fine-tuning and preference optimization pipelines and developed evaluation suites for layout validity and instruction-following accuracy.

  • Machine Learning Intern — Handshake AI (Joint with OpenAI)

    Aug 2025 – Nov 2025

    Leveraged a paper-to-code pipeline to convert ICLR papers into runnable PyTorch implementations; designed evaluation rubrics and graded LLM-generated code.

  • Research Assistant — UC Santa Barbara

    Sep 2023 – Present

    Developed low-rank and prompt-based adaptation methods for VLMs; designed communication-efficient federated fine-tuning algorithms using tensor decomposition; proposed decentralized and Byzantine-resilient optimization methods.

Publications

  • Preprint

    Can MLLMs Critique Like Humans? Evaluating Open-Ended Aesthetic Reasoning in Multimodal Large Language Models

    S. Ghiasvand, M. Amirizaniani, H. Ehsani Oskouie, M. Alizadeh, R. Pedarsani

    arXiv, 2026 · PDF

  • Preprint

    REALM: Reliable Expertise-Aware Language Model Fine-Tuning from Noisy Annotations

    S. Ghiasvand, M. Beliaev, M. Alizadeh, R. Pedarsani

    arXiv, 2026 · PDF

  • ECCV 2026

    MMLoP: Multi-Modal Low-Rank Prompting for Efficient Vision-Language Adaptation

    S. Ghiasvand, H. E. Oskouie, M. Alizadeh, R. Pedarsani

    ECCV 2026 · PDF

  • ICLR 2026

    pFedMMA: Personalized Federated Fine-Tuning with Multi-Modal Adapter for Vision-Language Models

    S. Ghiasvand, M. Alizadeh, R. Pedarsani

    ICLR 2026 · PDF

  • ICLR Workshop

    Few-Shot Adversarial Low-Rank Fine-Tuning of Vision-Language Models

    S. Ghiasvand, H. E. Oskouie, M. Alizadeh, R. Pedarsani

    Trustworthy AI @ ICLR 2026 · PDF

  • ACL 2025

    Communication-Efficient and Tensorized Federated Fine-Tuning of Large Language Models

    S. Ghiasvand, Y. Yang, Z. Xue, M. Alizadeh, Z. Zhang, R. Pedarsani

    Findings of ACL 2025 · PDF

  • ACL Workshop

    Decentralized Low-Rank Fine-Tuning of Large Language Models

    S. Ghiasvand, M. Alizadeh, R. Pedarsani

    REALM Workshop @ ACL 2025 · PDF

  • Journal

    Robust Decentralized Learning with Local Updates and Gradient Tracking

    S. Ghiasvand, A. Reisizadeh, M. Alizadeh, R. Pedarsani

    IEEE/ACM Transactions on Networking, 2025 · PDF

  • Allerton 2024

    Communication-efficient and Decentralized Federated Minimax Optimization

    S. Ghiasvand, A. Reisizadeh, M. Alizadeh, R. Pedarsani

    Allerton Conference, 2024 · PDF

  • Preprint

    Exploring Cross-model Neuronal Correlations in the Context of Predicting Model Performance and Generalizability

    H. E. Oskouie, S. Ghiasvand, L. Levine, M. Sarrafzadeh

    arXiv, 2024