Abdelmajid Essofi

I am a research assistant at Mohamed Bin Zayed University of Artificial Intelligence (MBZUAI) under the supervision of Dr. Qirong Ho and Dr. Willie Neiswanger. We are building a robust ecosystem which proposes sustainable solutions to complex optimization problems, such as tuning tasks and automated resource scheduling and allocation. Our work contributes to minimizing the carbon footprint incurred from training large scale AI models and pipelines through cost-efficient approaches. We aim to push the boundaries of cost-effectiveness and sustainability in large-scale AI infrastructure towards a net-zero emission objective.

My academic timeline.

Education

I am currently pursuing a PhD in machine learning under the esteemed supervision of Dr. Takáč In 2021-2022, I attended MBZUAI, where I defended my thesis to obtain my MSc in Machine Learning from MBZUAI under the supervision of Dr. Le Song.
In 2015-2019, I attended Al Akhawayn University in Ifrane (AUI), where I obtained my BSc in Computer Science under the supervision of Dr. Violetta Cavalli-Sforza.
In Fall 2017, I attended Beloit College as an exchange student.

Projects & Publications

Explaining Point Processes by Learning Interpretable Temporal Logic Rules
A principled method to learn a set of human-readable logic rules to explain temporal point processes. Work accepted in ICLR2022.

Memoization-aware Bayesian Optimization for Multistage AI Pipelines with Unknown Costs
A toolkit for cost-efficient hyperparameter tuning via memoization and uncertainty modeling. Work submmitted for ICLR2024.

Alpa-AdaptDL
Ongoing project for resource-adaptive inter and intra-cluster scheduling for deep learning training.

Recovering Lost Information from Exponential Moving Averages on HFT Market Data
Preventing EMA deviation from insignificant adjacent time differences. The proposed method was tested against the baseline by comparing trading profits of an RL agent using signals based on each method.

NLP For Learning Tamazight
A language model using system-user interaction to learn the grammatical structure of the Moroccan-native Tamazight language.

Kudos
Scheduling platform for tutoring sessions offered by Al Akhawayn University

Open Energy Dashboard
Open-source project by Beloit college for recording and visualizing power consumption levels for given organizations.

Programming Contests

In 2016, 2018, and 2019, I represented my university, along with two teammates, in the annual Moroccan Collegiate Programming contest (MCPC), where my team successfully obtained the first place in both MCPC 2018, and MCPC 2019. From the Moroccan regionals, we qualified to represent our country in the Arab Collegiate Programming Contest, where we ranked first among the teams of the Maghreb region in the 2018 edition to obtain the Maghreb Champions prize.

Internships

Jump Trading - Singapore
In June 2022, I was introduced to the incredible uses of optimal machine learning techniques for high frequency trading during my internship as a quantitative researcher in Jump Trading, where I built a linear model that uses pre-defined alpha signals to predict the future price change of a given symbol, as well as a reinforcement learning agent that performs trades based on the current market state, with the aim of maximizing the end-of-day profits.

Alicante Healthcare - France
After obtaining my bachelor’s degree in December 2019, I spent the first six month of the Covid era in Lille as a Big Data Analytics intern with Alicante, where I learned to use Spark and Hadoop to receive data from a streaming pipeline, run statistics on notes received by healthcare centers, before pushing the retrieved statistics to a cloud-based database.

Leyton - Morocco
In June 2019, I took my first steps into the field of AI by joining Leyton as a machine learning engineering intern, where I was in charge of building a prediction model that uses the data collected on advertisement panels to predict the prices of posting advertisements in each panel.

Peixoto Atacadista - Brazil
In March 2018, I joined Peixoto Atacadista, a logistics company in Uberlândia, for my first internship as a full-stack software development intern, where I developed a website for their human resources department to manage applications and filter candidate using given filters.