Maya Bechler-Speicher

AI Research Scientist

Meta Tel-Aviv University

I'm an AI Research Scientist at Meta, working mainly on Graph Foundation Models and LLMs post-training with structured data.

I am also a lecturer at the Computer Science School at Tel-Aviv University, teaching "Machine Learning with Graphs" — an advanced course I built from scratch to spread the word on Graph Machine Learning.

I defended my PhD "Towards improved Generalizability and Interpretability in Graph Neural Networks" at the School of Computer Science at Tel Aviv University, where I was fortunate to be advised by Amir Globerson.

I am broadly interested in Deep Learning and Geometric Deep Learning.

In the summer of 2021, I interned at Meta. Prior to that, I spent three years at Microsoft's Machine Learning Incubation and Innovation group (CTO office), where I had the opportunity to invent, lead, and develop novel and disruptive AI-based products.

I hold a BSc and MSc in Computer Science from Ben-Gurion University, where I conducted research with Natan Rubin and participated in the 'Dkalim' and 'Intel' excellence programs. I also completed a second full BSc in Mathematics after my MSc and mostly in parallel to my PhD — just for fun. I really love Math.

Scroll

News

2026

Invited Talk on Graph Foundation Models at the Workshop on Geometry-grounded Representation Learning and Generative Modeling (GRaM) at ICLR 2026.

2026

New preprint: 'Billion-Scale Graph Foundation Models' is now available!

2026

Two papers and one workshop paper accepted to ICLR 2026!

2025

New preprint on next-generation graph benchmarking: GraphBench — try it at graphbench.github.io!

2025

Two spotlight papers and two workshop papers accepted to NeurIPS 2025!

2025

Talk at the Center of Computational Mathematics at the FlatIron Institute, July 2025.

2025

Keynote Talk at GHOST Day — Applied Machine Learning conference.

2025

Paper "Position: Graph Learning Will Lose Relevance Due To Poor Benchmarks" accepted to ICML 2025!

2025

Speaking at Joan Bruna's ML seminar CS@NYU, March 2025.

Publications

Preprint

Lost in Tokenization: Fundamental Trade-offs in Graph Tokenization for Transformers

Maya Bechler-Speicher*, Gilad Yehudai*, Gil Harari, Clayton Sanford, Amir Globerson, Joan Bruna

Preprint

ACL 2026 Workshop

Ex-GraphRAG: Interpretable Evidence Routing for Graph-Augmented LLMs

Yoav Kor Sade*, Arvindh Arun*, Rishi Puri, Steffen Staab, Maya Bechler-Speicher

ACL 2026 Towards Knowledgeable Foundation Models (KnowFM) workshop

ICLR 2026 Workshop

Improving LLM Predictions via Inter-Layer Structural Encoders

Tom Ulanovski*, Eyal Blyachman*, Maya Bechler-Speicher

Workshop on Geometry-grounded Representation Learning and Generative Modeling (GRaM), ICLR 2026

Preprint

Billion-Scale Graph Foundation Models

Maya Bechler-Speicher, Yoel Gottlieb, Andrey Isakov, David Abensur, Ami Tavory, Daniel Haimovich, Ido Guy, Udi Weinsberg

Preprint

ICLR 2026

A Graph Meta-Network for Learning on Kolmogorov-Arnold Networks

Guy Bar-Shalom, Ami Tavory, Itay Evron, Maya Bechler-Speicher, Ido Guy, Haggai Maron

International Conference on Learning Representations (ICLR) 2026

ICML 2026 Workshop

GraphBench: Next-generation graph learning benchmarking

Timo Stoll, Chendi Qian, Ben Finkelshtein, Ali Parviz, Darius Weber, Fabrizio Frasca, Hadar Shavit, Antoine Siraudin, Arman Mielke, Marie Anastacio, Erik Müller, Maya Bechler-Speicher, Michael Bronstein, Mikhail Galkin, Holger Hoos, Mathias Niepert, Bryan Perozzi, Jan Tönshoff, Christopher Morris

ICML 2026 workshop on Graph Foundation Models Oral

ICLR 2026

SuperMAN: Interpretable and Expressive Networks over Temporally Sparse Heterogeneous Data

Maya Bechler-Speicher*, Andrea Zerio*, Maor Huri, Marie Vibeke Vestergaard, Ran Gilad-Bachrach, Tine Jess, Samir Bhatt, Aleksejs Sazonovs

International Conference on Learning Representations (ICLR) 2026

NeurIPS 2025 Workshop

Graph Mixing Additive Models

Maya Bechler-Speicher*, Andrea Zerio*, Maor Huri, Marie Vibeke Vestergaard, Ran Gilad-Bachrach, Tine Jess, Samir Bhatt, Aleksejs Sazonovs

NeurIPS Symmetry and Geometry in Neural Representations (NeurReps) Workshop 2025

NeurIPS 2025 Spotlight

Spectral Graph Neural Networks are Incomplete on Graphs with a Simple Spectrum

Snir Hordan, Maya Bechler-Speicher, Gur Lifshitz, Nadav Dym

Advances in Neural Information Processing Systems (NeurIPS) 2025 Spotlight

NeurIPS 2025 Spotlight

Depth-Width Tradeoffs in Algorithmic Reasoning of Graph Tasks with Transformers

Gilad Yehudai, Clayton Sanford, Maya Bechler-Speicher, Orr Fischer, Ran Gilad-Bachrach, Amir Globerson

Advances in Neural Information Processing Systems (NeurIPS) 2025 Spotlight

ICML 2025

Position: Graph Learning Will Lose Relevance Due To Poor Benchmarks

Maya Bechler-Speicher*, Ben Finkelshtein*, Fabrizio Frasca*, Luis Müller*, Jan Tönshoff*, Antoine Siraudin, Viktor Zaverkin, Michael M. Bronstein, Mathias Niepert, Bryan Perozzi, Mikhail Galkin, Christopher Morris

International Conference on Machine Learning (ICML) 2025

ICLR 2025 Workshop

Identifying Critical Phases for Disease Onset with Sparse Haematological Biomarkers

Andrea Zerio, Maya Bechler-Speicher, Tine Jess, Aleksejs Sazonovs

International Conference on Learning Representations (ICLR) 2025, LMRL Workshop

Preprint

Towards Invariance to Node Identifiers in Graph Neural Networks

Maya Bechler-Speicher, Moshe Eliasof, Carola-Bibiane Schönlieb, Ran Gilad-Bachrach, Amir Globerson

Preprint

A General Recipe for Contractive Graph Neural Networks — Technical Report

Maya Bechler-Speicher, Moshe Eliasof

Preprint

LOG 2024

Cayley Graph Propagation

JJ Wilson, Maya Bechler-Speicher, Petar Velickovic

Learning On Graphs (LOG) 2024

NeurIPS 2024

The Intelligible and Effective Graph Neural Additive Networks

Maya Bechler-Speicher, Amir Globerson, Ran Gilad-Bachrach

Advances in Neural Information Processing Systems 37 (NeurIPS) 2024

ICML 2024

Graph Neural Networks Use Graphs When They Shouldn't

Maya Bechler-Speicher, Ido Amos, Ran Gilad-Bachrach, Amir Globerson

International Conference on Machine Learning (ICML) 2024

AAAI 2024

TREE-G: Decision Trees Contesting Graph Neural Networks

Maya Bechler-Speicher, Amir Globerson, Ran Gilad-Bachrach

Association for the Advancement of Artificial Intelligence (AAAI) 2024

US Patent

System and method for improving machine learning models based on confusion error evaluation

Oren Elisha, Ami Luttwak, Hila Yehuda, Adar Kahana, Maya Bechler-Speicher

US Patent

System and method for improving machine learning models by detecting and removing inaccurate training data

Oren Elisha, Ami Luttwak, Hila Yehuda, Adar Kahana, Maya Bechler-Speicher

US Patent

Iterative vectoring for constructing data driven machine learning models

Oren Elisha, Ami Luttwak, Hila Yehuda, Adar Kahana, Maya Bechler-Speicher

US Patent

Mathematics

A Crossing Lemma for Families of Jordan Curves with a Bounded Intersection Number

Maya Bechler-Speicher, Natan Rubin

Industry

Research Scientist

Meta · Full-time

Nov 2024 – Present

Research Lead on Foundation Models for Graphs, LLMs post-training & Agents with Geometric Priors, and Security.

Founder & AI Consultant

Drop of Wisdom — AI Specialists

2021 – Present

Visiting Researcher

University of Cambridge · DAMTP

2024

Visiting researcher at the Department of Applied Mathematics and Theoretical Physics (DAMTP), hosted by Professor Carola Bibiane Schönlieb.

Data & Applied Scientist

Microsoft · AI Incubation, CTO Office

Aug 2018 – Dec 2020

Worked in the AI Incubation team within the CTO office. Invented, led, and developed novel and disruptive AI-based products — taking ideas from research prototypes to production-grade systems.

Software Engineer

Microsoft · Innovation & Incubations, CTO Office

Nov 2017 – Aug 2018

Part of the Innovation & Incubations team in the CTO office, developing cutting-edge software solutions and early-stage AI products.

Software Engineer

Autodesk · AutoCAD Mobile Group

Aug 2016 – Nov 2017

Intern at Google Camp

Google · London, UK

Aug – Dec 2016

Selected for a unique internship camp at Google London for 25 qualified computer science students from around the world.

Talks & Invited Presentations

2026

Invited Talk — GRaM Workshop, ICLR 2026

Graph Foundation Models at the Workshop on Geometry-grounded Representation Learning and Generative Modeling.

2025

Keynote Talk — GHOST Day

Applied Machine Learning conference keynote.

2025

ML Seminar — CS@NYU

Speaking at Joan Bruna's ML seminar, March 2025.

2025

FlatIron Institute — Center of Computational Mathematics

Talk at the Center of Computational Mathematics, July 2025.

2024

Haggai Maron's Group Seminar — Technion

Talk on implicit biases in Graph Neural Networks.

2024

Chaim Baskin's Group Seminar — BGU

Talk on implicit biases in Graph Neural Networks.

2024

Graph Learning on Wednesdays (GLOW)

Talk on "Graph Neural Networks Use Graphs When They Shouldn't", 18/11/24.

2024

Geometric Deep Learning Tel-Aviv Meetup 2024

Co-organized as part of LOG 2024 local meetups.

2024

Cambridge Image and Analysis Seminar

Department of Applied Mathematics, University of Cambridge. Talk on "Graph Neural Networks Use Graphs When They Shouldn't". (YouTube)

2024

ML Theory Summer School — Princeton University

Participant at the Machine Learning Theory Summer School.

2023

Deep Learning Theory Retreat

Talk at the AI and Data Science Center retreat.

2022

Women in Theory (WIT) — Simons Institute, Berkeley

Talk and participation at the Women in Theory Conference at the Simons Institute for the Theory of Computing.

Maya Bechler-Speicher

News

Publications

Industry

Research Scientist

Founder & AI Consultant

Visiting Researcher

Data & Applied Scientist

Software Engineer

Software Engineer

Intern at Google Camp

Talks & Invited Presentations

Invited Talk — GRaM Workshop, ICLR 2026

ML Seminar — CS@NYU

FlatIron Institute — Center of Computational Mathematics

Haggai Maron's Group Seminar — Technion

Chaim Baskin's Group Seminar — BGU

Geometric Deep Learning Tel-Aviv Meetup 2024

Cambridge Image and Analysis Seminar

ML Theory Summer School — Princeton University

Deep Learning Theory Retreat

Women in Theory (WIT) — Simons Institute, Berkeley

Teaching

Machine Learning with Graphs

Introduction to Data Science

Algorithms in Geometric Networks

Geometric Algorithms

Systems Programming

Reviewer Service