CV
General Information
Full Name | Amir Joudaki |
Location | Zurich, SWITZERLAND |
Languages | English, Persian, German (A1) |
Education
-
March 2019 - now
Direct Ph.D. in Artificial Intelligence
ETH Zurich
- Focused on mathematical foundations of AI and AI for biomedicine.
- MSc GPA. 5.68/6 (top 10%).
-
Feb 2014 - Feb 2017
M.Sc. in Cognitive Neuroscience
International School for Advanced Studies (SISSA)
- Thesis on electrophysiological recordings in vivo in rats.
-
Sept 2008 - Sept 2011
B.Sc. in Computer Engineering
Sharif University of Technology
Experience
-
Jan 2022 - July 2022
Machine Learning Consultant
SkillLab
- Developed a proof-of-concept for training, labeling, and monitoring NLP tasks using PyTorch, Flask, Typescript, React, MongoDB; orchestrated as microservices in REST API and deployed using Docker containers.
-
Jan 2013 - Dec 2013
Research Assistant
Sharif University of Technology
- Conducted network-based analysis of EEG recordings for control and patient groups using MATLAB.
-
Sept 2011 - Dec 2012
Software Engineer Intern
Rahtech Hooshmand Sharif
- Developed a vehicle tracking and monitoring product using Qt/C++.
Awards
-
2017
- Direct Doctorate Fellowship - Among two out of more than one hundred students selected for direct PhD program at ETH computer science department
-
2011
- Ranked 42nd - Among more than 50,000 participants in National Higher Education Entrance Exam
-
2007
- Ranked 369th - Among more than 400,000 participants in National University Entrance Exam
Projects
-
Human-in-the-Loop Pipeline for Occupation and Skill Extraction
- Served as AI/NLP consultant, focusing on extracting occupations and skills from unstructured text and normalizing them to European Skills, Competences, and Occupations (ESCO) standards.
- Designed a human-in-the-loop pipeline to validate model predictions, implemented as microservices deployed with Docker Compose.
Publications
-
Mathematical Foundations of AI
- Batch normalization without gradient explosion: Towards training without depth limits, ICLR 2024.
- On the impact of activation and normalization in obtaining isometric embeddings at initialization, NeurIPS 2023
- On Bridging the Gap between Mean Field and Finite Width in Deep Random Neural Networks with Batch Normalization, ICML 2023
- Batch Normalization Orthogonalizes Representations in Deep Random Networks, spotlighted NeurIPS 2021 (top 3% of submissions)
-
Genomic Sequence Analysis
- Learning Genomic Sequence Representations using Graph Neural Networks over De Bruijn Graphs, NeurIPS workshop 2023
- Aligning distant sequences to graphs using long seed sketches, Journal of Genome Research 2023
-
Dimensionality Reduction
- Nonlinear Dimensionality Reduction via Path-Based Isometric Mapping, IEEE transactions on pattern analysis and machine intelligence
-
Functional Brain Networks Analysis
- Properties of functional brain networks affect frequency of psychogenic non-epileptic seizures, Frontiers in Human Neuroscience
- EEG-based functional brain networks: does the network size matter?, PLoS ONE