Ali Zeynali

About

Ph.D. student in Computer Science at UMass Amherst, advised by Prof. Mohammad Hajiesmaili and Prof. Ramesh K. Sitaraman.

I design robust learning and online decision-making algorithms for large-scale ML systems, with emphasis on data-efficient fine-tuning under limited or noisy data.

CV Google Scholar GitHub LinkedIn

Research

Online learning, decision-making, and trustworthy ML with a focus on robust, learning-augmented methods.

Engineering

I had industry experience at Google, Adobe, Airbnb, and SiriusXM/Pandora working on cutting-edge technologies and scalable AI models.

Interests

Scalable ML systems, fine-tuning under noisy/limited data, and practical deployment.

Industry Experience

• Machine Learning Data Scientist Intern, Airbnb (Remote), Summer/Fall 2025
  • Built a framework to detect synthetic/manipulated images.
  • Designed methods to explain fake-image detection predictions.
• Machine Learning Data Scientist Intern, SiriusXM and Pandora (Oakland, USA), Summer 2024
  • Built an explainability framework for search ranking relevance.
  • Improved prior model performance by up to 26%.
• Software Engineer Intern, Google (Mountain View, USA), Spring 2022
  • Improved interest-matching accuracy with geometry/depth-aware vision methods.
  • Improved over prior SOTA models by up to 50%.
• ML/Data Science Research Intern, Adobe (San Jose, USA), Summer 2021
  • Improved Adobe Liquid Mode streaming model under online/offline settings.
  • Improved quality of experience and bandwidth efficiency by up to 20%.

Selected Publications

Near-Optimal Emission-Aware Online Ride Assignment Algorithm for Peak Demand Hours

Venue: Under review · PDF

Mahsa Sahebdel, Ali Zeynali, Noman Bashir, Ramesh Sitaraman, Mohammad Hajiesmaili

Smoothed Online Optimization for Target Tracking: Robust and Learning-Augmented Algorithms

Venue: ALT 2026 · PDF

Ali Zeynali, Mahsa Sahebdel, Qingong Liu, Ramesh Sitaraman, Mohammad Hajiesmaili

LEAD: Towards Learning-Based Equity-Aware Decarbonization in Ridesharing Platforms

Venue: ACM FAccT 2025 · PDF

Mahsa Sahebdel, Ali Zeynali, Noman Bashir, Prashant Shenoy, Mohammad Hajiesmaili

BOLA360: Near-optimal View and Bitrate Adaptation for 360-degree Video Streaming - Journal Version

Venue: ACM TOMM 2025 · PDF

Ali Zeynali, Mahsa Sahebdel, Mohammad Hajiesmaili, Ramesh Sitaraman

Robust Learning-Augmented Dictionaries

Venue: ICML 2024 · PDF

Ali Zeynali, Shahin Kamali, Mohammad Hajiesmaili

BOLA360: Near-optimal View and Bitrate Adaptation for 360-degree Video Streaming

Venue: ACM MMSys 2024 · PDF

Ali Zeynali, Mohammad Hajiesmaili, Ramesh Sitaraman

A Holistic Approach for Equity-aware Carbon Reduction of Ridesharing Platforms

Venue: ACM e-Energy 2024 · PDF

Mahsa Sahebdel, Ali Zeynali, Noman Bashir, Prashant Shenoy, Mohammad Hajiesmaili

Data-driven Algorithms for Reducing the Carbon Footprint of Ride-sharing Ecosystems

Venue: ACM e-Energy 2023 · PDF

Mahsa Sahebdel, Ali Zeynali, Noman Bashir, Mohammad Hajiesmaili, Jimi Oke

SemEval-2022 Task 8: Multilingual News Article Similarity

Venue: SemEval 2022 · PDF

Xi Chen, Ali Zeynali, Chico Camargo, Fabian Flock, Devin Gaffney, Przemyslaw Grabowicz, Scott Hale, David Jurgens, Mattia Samory

Competitive Algorithms for Online Multidimensional Knapsack Problems

Venue: ACM SIGMETRICS 2022 · PDF

Lin Yang, Ali Zeynali, Mohammad Hajiesmaili, Ramesh Sitaraman, Don Towsley

Data-driven Competitive Algorithms for Online Knapsack and Set Cover

Venue: AAAI 2021 · PDF

Ali Zeynali, Bo Sun, Mohammad Hajiesmaili, Adam Wierman

Competitive Algorithms for the Online Multiple Knapsack Problem with Application to Electric Vehicle Charging

Venue: ACM SIGMETRICS 2021 · PDF

Bo Sun, Ali Zeynali, Tongxin Li, Mohammad Hajiesmaili, Adam Wierman, Danny H.K. Tsang

Projects

Robust and Unbiased Self-Play Strategy for Fine-Tuning Large Language Models

Framework for fine-tuning large language models on imbalanced datasets with robust self-play strategies.

LLM Fine-Tuning Robust Learning

VSE360: Online 360-degree video streaming simulation environment

Fully simulated Python environment for evaluating 360-degree ABR algorithms.

Video Streaming Robust Learning GitHub

AI-Generated Music using Deep Learning + LSTM

Music generation pipeline based on deep sequence models and LSTM architectures.

LSTM AI Training Time Series GitHub

AI-Generated Short Stories using Bidirectional LSTM

Tiny-story generation model powered by deep bidirectional LSTM networks.

Bi-LSTM AI Training Time Series GitHub

ZeySed: Deep neural networks for leaf classification

Image classification pipeline for plant leaf datasets using deep neural networks.

CNN Image Processing GitHub

Honors and Awards

• Thesis Proposal Writing Fellowship Award, UMass Amherst (Fall 2023)
• Nominated for Microsoft Research Fellowship by CICS Department, UMass Amherst (Summer 2021)
• Donald F. Towsley Graduate Scholarship, UMass Amherst (Summer 2021)
• Top-10 B.Sc. students of Computer Engineering class of 2019 (among 177 students)
• Rank 24, IEEEXtreme 10th 24-hour programming contest (among 1823 teams)
• Gold Medal, 8th International Olympiad on Astronomy and Astrophysics (IOAA), 2014
• Gold Medal, 9th National Science Olympiad (NOAA), 2013