About Me

I'm currently a senior researcher and manager at Apple, leading a team of engineers and scientists to enable efficient inference of Large Language Models (LLMs) under resource constraints. Besides efficient inference, I work on understanding and demystifying how large vision and language models work and learn, in order to find more accurate and efficient pretraining and finetuning architectures, algorithms, and strategies. I also work on  vision language foundation models (e.g., CLIP) and lead efforts to build continual, multi-task, and multi-modal learning algorithms that accumulate knowledge and skills over time. Additionally, I study how to efficiently transfer their knowledge to smaller models or to downstream and future unseen tasks.

Before that, I was a senior research scientist at DeepMind. As a research scientist, I worked on continual and lifelong learning, multitask and transfer learning, understanding the training dynamics of deep neural networks, and reinforcement learning. These research areas were in line with DeepMind's mission towards Artificial General Intelligence. As an applied scientist and engineer, I worked on applications of machine learning, e.g., using recommendation/predictive models, meta learning, causal inference, and reinforcement learning to improve Google's products in areas such as YouTube, Cloud, and Sales.

I received my Ph.D. in Computational Science and Engineering from Georgia Institute of Technology, under the supervision of Hongyuan Zha and Le Song. I used to work on modeling and optimization of sequential events data, stochastic point processes, and dynamics of and on the networks. During my PhD I interned at Micorosft Research, Max Planck Instintute for Software Systems and Google, working on predicting and leveraging health data, information reliability, and analyzing google maps local listings data. I received my M.Sc. in Artificial Intelligence from the Computer Engineering Department at Sharif University of Technology and my B.Sc. from the same university in Software Engineering, in 2011 and 2009, respectively.

Google Scholar LinkedIn Twitter

Prospective Candidates: Our group may have openings for exceptional candidates on LLM reasoning and LLM inference efficiency! If you have prior hands-on experience and proven track of related research papers at top conferences on the above topics don't hesitate to reach out via my last name at apple dot com!