About Me

Hi! 👋 I’m a 1st year Masters in CS student at Harvard University with an interest in deep learning, computer vision & natural language processing.

I am a research assistant focused on self-supervised representation Learning, vision-language learning, and adversarial robustness. Previously I worked in deep learning for healthcare at Stanford and also spent some time at Twitch/Amazon as an Applied Scientist where I worked on self-supervised and continual visual representation learning on streaming video data.

Experience

Twitch/Amazon 3X Internship 📺

  • 2022 (Applied Science): Developed unsupervised, continual learning framework on Twitch Streaming data. Framework allows for ML models developed by Twitch to continuously and intelligently update on new streams such that a model learns new representations without forgetting old ones.

  • 2021 Fall (Software Engineer): Designed & developed an end-to-end, real-time Media-Analysis backend service integrated with both other AWS services and my trained ML model to moderate Twitch’s 9.36M+ streamers’ content. Service is in production!

  • 2021 Summer (Applied Science): Developed a Twitch content representation Image Embedding trained using self-supervised learning. Improved prior method’s performance on downstream tasks such as Game Stream Classification by ~10%.

Undergraduate Research Intern đź’»

  • Research @ UC Davis under Dr. Hamed Pirsiavash

    • Project 1: Adversarial robustness for vision transformers
    • Project 2: Improving compositional reasoning of vision-language models
  • Research @ Stanford University under Dr. Dennis Wall and Dr. Peter Washington.

    • Project 1 (First Author): Deep Learning-Based Autism Spectrum Disorder (ASD) Detection Using Emotion Features From Video Recordings.
      [Published in JMIR 2022].
    • Project 2: Novel Facial Emotion Recognition Dataset: “TikTok for Good: Creating a Diverse Emotion Expression Database”.
      [Published at CVPRW 2022].
    • Project 3: Mitigating Negative Transfer in Multi-Task Learning with Exponential Moving Average Loss Weighting Strategies.
      [Published at AAAI Student Abstract 2023]

Publications

SlowFormer: Universal Adversarial Patch for Attack on Compute and Energy Efficiency of Inference Efficient Vision Transformers
KL Navaneet*, Soroush Abbasi Koohpayegani*, Essam Sleiman*, Hamed Pirsiavash
arXiv. [link]

Deep Learning-Based Autism Spectrum Disorder Detection Using Emotion Features From Video Recordings
Essam Sleiman; Onur Cezmi Mutlu; Saimourya Surabhi; Arman Husic; Aaron Kline; Peter Washington; Dennis P. Wall
JMIR 2022. [pdf]

TikTok for Good: Creating a Diverse Emotion Expression Database
Saimourya Surabhi,…,Essam Sleiman, Dennis P. Wall
Computer Vision and Pattern Recognition (CVPR) Workshop, 2022. [pdf]

Mitigating Negative Transfer in Multi-Task Learning with Exponential Moving Average Loss Weighting Strategies.
Anish Lakkapragada, Essam Sleiman, Mourya Surabhi, Dennis P. Wall
AAAI 2023 (Student Abstract)

Continually Learning Self-Supervised Image Embeddings
Essam Sleiman, Xiangbo Li, Saad Ali, Lukas Tencer
2022 [Processing: Email me for access to pdf.]