Essam Sleiman
I run Canvas Labs (YC), where we work on self-improving AI.
Previously, I did research on continual learning, language-model interpretability, and vision-language models at Amazon Science and as a grad student at Harvard, with papers at ECCV and CVPR. In my free time, I enjoy basketball, coffee, and hiking around SF.
Selected Work
Canvas Labs
AutoHarnessBench
A long-horizon SWE & self-improvement benchmark evaluating automatic harness optimization.
Canvas Labs
Privileged Self-Distillation
A post-training method that distills training signal from failed rollouts corrected with privileged hints.
Research Engineer at Amazon Science
Continual Self-Supervised Learning with Knowledge Distillation
A video embedding model for Twitch that learns new content without forgetting the old.
Research, Harvard
Goldfish: Vision-Language Understanding of Arbitrarily Long Videos
A SOTA vision-language model for question answering on hour-long videos.
Research, Harvard & MIT
Do Function Vectors Factor Task and Distribution?
An interpretability study of how LLMs represent tasks during in-context learning.
Research, UC Davis
SlowFormer: Adversarial Robustness for Efficient Vision Transformers
A universal patch attack that maximizes the compute cost of efficient vision transformers.