Itzik Malkiel

I'm a Principal Researcher at Microsoft, working on projects in Dynamics 365 for Commerce and Intelligent Recommendations. Prior to joining Microsoft, I was a research scientist at GE Global Research and a senior software engineer and project leader. I hold Ph.D. and M.Sc. degrees in Computer Science from Tel-Aviv University, supervised by Prof. Lior Wolf, and a B.Sc. degree in Computer Science from the Hebrew University of Jerusalem.

My current research focuses on machine learning and deep learning and includes topics such as natural language understanding, healthcare and medical imaging, recommender systems, computer vision, representation learning, and inverse problems.

Email / Google Scholar / Github / LinkedIn

Research

Representative papers are highlighted.

	SEGLLM: Topic-Oriented Call Segmentation Via LLM-Based Conversation Synthesis I Malkiel, U Alon, Y Yehuda, S Keren, O Barkan, R Ronen, N Koenigstein. IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP), 2024.
	Unsupervised Topic-Conditional Extractive Summarization I Malkiel, Y Yehuda, J Ephrath, O Kats, O Barkan, N Nice, N Koenigstein. IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP), 2024.
	Efficient Discovery and Effective Evaluation of Visual Perceptual Similarity: A Benchmark and Beyond O. Barkan, T. Reiss, J. Weill, O. Katz, R. Hirsch,I. Malkiel, N. Koenigstein. International Conference on Computer Vision (ICCV), 2023.
	Harnessing GPT for Topic-Based Call Segmentation in Microsoft Dynamics 365 Sales I Malkiel, U Alon, Y Yehuda, S Keren, N Koenigstein International Conference on Information and Knowledge Management (CIKM), Industry Track, 2023.
	Pre-Training Transformers for Fingerprinting to Improve Stress Prediction in fMRI G Rosenman, I Malkiel, A Greental, T Hendler, L Wolf Medical Imaging with Deep Learning (MIDL), 2023. Long paper. github
	GPT-Distilled Calls Segmentation and Tagging for Conversation Intelligence in Dynamics 365 Sales I Malkiel, U Alon, Y Yehuda, S Keren, N Koenigstein Machine Learning, AI and Data Science Conference (MLADS), 2023. Microsoft Journal of Applied Research (MSJAR), 2023.
	Intelligent Recommendations: Self-Supervised Triplet Training for Textual Item Similarity I Malkiel, D Ginzburg, O Barkan, A Caciularu, Y Weill, O Katz, N Koenigstein Machine Learning, AI and Data Science Conference (MLADS), 2022. Microsoft Journal of Applied Research (MSJAR), 2022.
	Self-Supervised Transformers for fMRI representation I Malkiel, G Rosenman, L Wolf, T Hendler Medical Imaging with Deep Learning (MIDL), 2022. Long paper.
	MetricBERT: Text Representation Learning Via Self-Supervised Triplet Training I Malkiel, D Ginzburg, O Barkan, A Caciularu, J Weill, N Koenigstein IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP), 2022.
	Interpreting BERT-based Text Similarity via Activation and Saliency Maps I Malkiel, D Ginzburg, O Barkan, A Caciularu, J Weill, N Koenigstein The Web Conference (WWW), 2022. Long paper.
	MTAdam: Automatic Balancing of Multiple Training Loss Terms I Malkiel, L Wolf Empirical Methods in Natural Language Processing (EMNLP), 2021. Long paper.
	Caption Enriched Samples for Improving Hateful Memes Detection E Blaier, I Malkiel, L Wolf Empirical Methods in Natural Language Processing (EMNLP), 2021. Short paper. github
	GAM: Explainable Visual Similarity and Classification via Gradient Activation Maps O Barkan, O Armstrong, A Hertz, A Caciularu, O Katz, I Malkiel, N Koenigstein The 30th ACM International Conference on Information and Knowledge Management (CIKM), 2021. Long paper.
	Grad-SAM: Explaining Transformers via Gradient Self-Attention Maps O Barkan, E Hauon, A Caciularu, O Katz, I Malkiel, O Armstrong, N Koenigstein The 30th ACM International Conference on Information and Knowledge Management (CIKM), 2021. Short paper.
	Representation Learning via Variational Bayesian Networks O Barkan, A Caciularu, I Rejwan, O Katz, J Weill, I Malkiel, N Koenigstein The 30th ACM International Conference on Information and Knowledge Management (CIKM), 2021. Long paper.
	Self-Supervised Document Similarity Ranking via Contextualized Language Models and Hierarchical Inference D Ginzburg, I Malkiel, O Barkan, A Caciularu, N Koenigstein The Association for Computational Linguistics (ACL): Findings, 2021. Long paper.
	Inverse design of unparametrized nanostructures by generating images from spectra I Malkiel, M Mrejen, L Wolf, H Suchowski Optics Letters, 2021.
	Adaptive Gradient Balancing for Undersampled MRI Reconstruction and Image-to-Image Translation I Malkiel, S Ahn, V Taviani, A Menini, L Wolf, CJ Hardy International Conference on Computational Photography (ICCP), 2021. github
	Maximal Multiverse Learning for Promoting Cross-Task Generalization of Fine-Tuned Language Models I Malkiel, L Wolf European Chapter of the Association for Computational Linguistics (EACL), 2021. Long paper.
	Cold Item Recommendations via Hierarchical Item2vec O Barkan, A Caciularu, I Rejwan, O Katz, J Weill, I Malkiel, N Koenigstein IEEE International Conference on Data Mining (ICDM), 912-917, 2020.
	RecoBERT: A Catalog Language Model for Text-Based Recommendations I Malkiel, O Barkan, A Caciularu, N Razin, O Katz, N Koenigstein Empirical Methods in Natural Language Processing (EMNLP): Findings, 2020. Long paper.
	Retrieving Nanostructure Images from Spectra M Mrejen, I Malkiel, L Wolf, H Suchowski Conference on Lasers and Electro-Optics (CLEO), 2020.
	Scalable Attentive Sentence Pair Modeling via Distilled Sentence Embedding O Barkan, N Razin, I Malkiel, O Katz, A Caciularu, N Koenigstein AAAI Conference on Artificial Intelligence (AAAI), 2020.
	Machine Learning for Nanophotonics I Malkiel, M Mrejen, L Wolf, H Suchowski MRS Bulletin, 45(3), 221-229. 2020.
	Deep Learning for Design and Retrieval of Plasmonic Nanostructures M Mrejen, I Malkiel, A Nagler, U Arieli, L Wolf, H Suchowski Conference on Lasers and Electro-Optics (CLEO), 2019. 15th International Work-Conference on Artificial Neural Networks (IWANN), 2019.
	Towards Motion-Robust MRI – Autonomous Motion Timing and Correction during MR Scanning using Multi-Coil Data and a Deep-Learning Neural Network R Brada, M Rotman, R Wein, S Ahn, I Malkiel, CJ Hardy International Society for Magnetic Resonance in Medicine (ISMRM), 2019.
	Leveraging Conditional GANs with Adaptive Loss Balancing for MRI Sparse Reconstruction I Malkiel, S Ahn, V Taviani, A Menini, Z Slavens, CJ Hardy International Society for Magnetic Resonance in Medicine (ISMRM), 2019.
	Residual Network with Null Data Consistency for Sparse MRI Reconstruction CJ Hardy, I Malkiel, S Ahn, V Taviani, A Menini, Z Slavens ISMRM Workshop on Machine Learning, 2018.
	Improving Variable-Density Single-Shot Fast Spin Echo with Deep-Learning Reconstruction Using Variational Networks F Chen, V Taviani, I Malkiel, JY Cheng, J Shaikh, S Chang, CJ Hardy, JM Pauly, SS Vasanawala International Society for Magnetic Resonance in Medicine (ISMRM), 2018.
	Densely Connected Iterative Network for Sparse MRI Reconstruction I Malkiel, S Ahn, Z Slavens, V Taviani, CJ Hardy International Society for Magnetic Resonance in Medicine (ISMRM), 2018.
	Variable-Density Single-Shot Fast Spin Echo MR Imaging with Deep-Learning Reconstruction Using Variational Networks F Chen, V Taviani, I Malkiel, JY Cheng, J Shaikh, S Chang, CJ Hardy, JM Pauly, S Vasanawala Radiology, 2018. Collaboration with Stanford Health Care
	Plasmonic Nano-structures Design and Characterization via Deep Learning I Malkiel, M Mrejen, A Nagler, U Arieli, L Wolf, H Suchowski Light: Science & Applications, 2018.
	Deep Learning for the Design of Nano-photonic Structures I Malkiel, A Nagler, M Mrejen, U Arieli, L Wolf, H Suchowski International Conference on Computational Photography (ICCP), 2018. github

Patents

I. Malkiel, Y. Yehuda, J. Ephrath, O. Barkan, N. Nice, N. Koenigstein. Unsupervised Topic-Conditional Extractive Summarization. PCT (2024).
I. Malkiel, Y. Yehuda, O. Barkan, J. Weill, N. Koenigstein. Hallucination Detection System and Method for LLM-Generated Responses. PCT (2024).
O Barkan, O Armstrong, A Hertz, A Caciularu, O Katz, I Malkiel, N Koenigstein, N. Nice. Deep gradient activation map model refinement. US Patent App. 17/452,961, 2023.
O Barkan, A Caciularu, I Rejwan, Y Weill, N Koenigstein, O Katz, I Malkiel, N. Nice. Representation learning with side information. US Patent App. 17/452,818, 2023.
O. Barkan, E. Hauon, O. Katz, A. Caciularu, I Malkiel, O. Armstrong, A. Hertz, N. Koenigstein, N. Nice. Deep self-attention matrix model refinement. US Patent App. 17/452,952, 2023.
I Malkiel, D. Ginzburg, N. Koenigstein, O. Barkan, N. Nice. Self-supervised document-to-document similarity system. US Patent 11,580,764, 2023.
I Malkiel, N Koenigstein, O Barkan, D Ginzburg, N Nice. INTERPRETING TEXT-BASED SIMILARITY. US Patent App. 17/218,136, 2022.
I Malkiel, N. Koenigstein, O. Barkan, J. Ephrath, Y. Weill, N. Nice. SYSTEMS AND METHODS FOR SEMANTIC SEARCH VIA FOCUSED SUMMARIZATIONS. PCT (2022).
I Malkiel, N. Koenigstein, Y. Weill, O. Barkan, J. Ephrath, N. Nice. Deep Angular Similarity Learning. PCT (2022).
N. Koenigstein, O. Barkan, E. Hauon, A. Caciularu, O. Katz, I Malkiel, N. Nice, O. Armstrong, A. Hertz. Grad-SAM: Explaining Transformers via Gradient Self-Attention Maps. PCT (2021).
N. Koenigstein, O. Barkan, A. Caciularu, O. Katz, Y. Weill, I Malkiel, N. Nice, I. Rejwan. Representation Learning via Variational Bayesian Networks. PCT (2021).
I Malkiel, P. Roit, N. Koenigstein, N. Nice, O. Barkan. Text-based similarity system for cold start recommendations. US Patent App. 16/789,381, 2021.
I. Jansen, S. Ahn, C. J. Hardy, I Malkiel, R. Brada, M. Rotman. MRI system and method using neural network for detection of patient motion. US Patent 10,996,306, 2021.
I Malkiel, D. Ginzburg, N. Koenigstein, O. Barkan, N. Nice. Novel Documents Similarity System to Mitigate the Cold Start Problem in Documents Recommendations. PCT (2020).
I Malkiel, N. Koenigstein, O. Barkan, D. Ginzburg, N. Nice. Novel Explainability Method for Interpreting Text-Based Item Recommendations. PCT (2020).
C. J. Hardy, I Malkiel. SYSTEM AND METHOD FOR SPARSE IMAGE RECONSTRUCTION UTILIZING NULL DATA CONSISTENCY. US Patent App. 16/150,079, 2020.
M. Rotman, R. Brada, S. Ahn, C. J. Hardy, I Malkiel, R. Wein. SYSTEM AND METHOD FOR CORRECTING FOR PATIENT MOTION DURING MR SCANNING. US Patent App. 16/394,774, 2020.
R. Brada, C. J. Hardy, S. Ahn, I. Jansen, I Malkiel, M. Rotman, R. Wein. MRI system and method for detection and correction of patient motion. US Patent 10,806,370, 2020.
I Malkiel, H. J. Hardy. METHOD AND APPARATUS FOR USING GENERATIVE ADVERSARIAL NETWORKS IN MAGNETIC RESONANCE IMAGE RECONSTRUCTION. US Patent App. 16/276,135, 2020.
I Malkiel, A. Sangtae, H. J. Hardy. SYSTEM AND METHOD FOR SPARSE IMAGE RECONSTRUCTION. US Patent 20,190,266,761, 2019.
I Malkiel, A. Nagler, M. Mrejen, U. Arieli, L. Wolf, H. Suchowski. METHOD AND SYSTEM FOR CHARACTERIZING A NANOSTRUCTURE BY MACHINE LEARNING. IL Patent WO/2018/146,683, 2018.

HONORS & AWARDS

2024 Best Paper at "Microsoft Journal of Applied Research (MSJAR)".
Honor issuer: Microsoft.
2023 Best Paper at "Microsoft Journal of Applied Research (MSJAR)".
Honor issuer: Microsoft.
2019 Top Cited & Downloaded Papers of "Light: Science & Applications".
Honor issuer: Springer Nature.
2018 Impact Award - Empower and Inspire Each Other.
Honor issuer: GE Global Research, 2018.
2018 Award Winning Research - ISMRM Honors 2018.
Honor issuer: Stanford University.
2010 Dean's award for excellent students.
Honor issuer: The Hebrew University of Jerusalem.
2008 Dean's award for excellent students.
Honor issuer: The Hebrew University of Jerusalem.

Professional Activities

Conferences

2018-Today Programme Committee/Reviewer for IEEE Computer Society Conference on Computer Vision and Pattern Recognition (CVPR), International Conference on Computational Photography (ICCP), Optical Materials Express, The Medical Image Computing and Computer Assisted Intervention Society (MICCAI), European Chapter of the Association for Computational Linguistics (EACL), North American Chapter of the Association for Computational Linguistics (NAACL), Knowledge Discovery in Databases (KDD), Association for Computational Linguistics (ACL), Empirical Methods in Natural Language Processing (EMNLP), Association for the Advancement of Artificial Intelligence (AAAI).

Itzik Malkiel

Research

SEGLLM: Topic-Oriented Call Segmentation Via LLM-Based Conversation Synthesis

Unsupervised Topic-Conditional Extractive Summarization

Efficient Discovery and Effective Evaluation of Visual Perceptual Similarity: A Benchmark and Beyond

Harnessing GPT for Topic-Based Call Segmentation in Microsoft Dynamics 365 Sales

Pre-Training Transformers for Fingerprinting to Improve Stress Prediction in fMRI

GPT-Distilled Calls Segmentation and Tagging for Conversation Intelligence in Dynamics 365 Sales

Intelligent Recommendations: Self-Supervised Triplet Training for Textual Item Similarity

Self-Supervised Transformers for fMRI representation

MetricBERT: Text Representation Learning Via Self-Supervised Triplet Training

Interpreting BERT-based Text Similarity via Activation and Saliency Maps

MTAdam: Automatic Balancing of Multiple Training Loss Terms

Caption Enriched Samples for Improving Hateful Memes Detection

GAM: Explainable Visual Similarity and Classification via Gradient Activation Maps

Grad-SAM: Explaining Transformers via Gradient Self-Attention Maps

Representation Learning via Variational Bayesian Networks

Self-Supervised Document Similarity Ranking via Contextualized Language Models and Hierarchical Inference

Inverse design of unparametrized nanostructures by generating images from spectra

Adaptive Gradient Balancing for Undersampled MRI Reconstruction and Image-to-Image Translation

Maximal Multiverse Learning for Promoting Cross-Task Generalization of Fine-Tuned Language Models

Cold Item Recommendations via Hierarchical Item2vec

RecoBERT: A Catalog Language Model for Text-Based Recommendations

Retrieving Nanostructure Images from Spectra

Scalable Attentive Sentence Pair Modeling via Distilled Sentence Embedding

Machine Learning for Nanophotonics

Deep Learning for Design and Retrieval of Plasmonic Nanostructures

Towards Motion-Robust MRI – Autonomous Motion Timing and Correction during MR Scanning using Multi-Coil Data and a Deep-Learning Neural Network

Leveraging Conditional GANs with Adaptive Loss Balancing for MRI Sparse Reconstruction

Residual Network with Null Data Consistency for Sparse MRI Reconstruction

Improving Variable-Density Single-Shot Fast Spin Echo with Deep-Learning Reconstruction Using Variational Networks

Densely Connected Iterative Network for Sparse MRI Reconstruction

Variable-Density Single-Shot Fast Spin Echo MR Imaging with Deep-Learning Reconstruction Using Variational Networks

Plasmonic Nano-structures Design and Characterization via Deep Learning

Deep Learning for the Design of Nano-photonic Structures

Patents

HONORS & AWARDS

Professional Activities

Conferences