Moshe Mash

Moshe Mash, Ph.D.

I work as an AI Researcher at Dicta which develops AI algorithms and large language models and conducts fine-tuning for various NLP tasks such as automated punctuation, authorship embedding, etc.

Before that, I completed a postdoctoral fellowship in the Reliable Autonomous Systems Lab at Carnegie-Mellon University Robotics Institute. My advisors were Prof. Reid Simmons and Prof. Stephanie Rosenthal.

I possess a Ph.D. from the Department of Software and Information Systems Engineering at Ben-Gurion University, which was supervised by Prof. Kobi Gal of the AI and Data Science Lab.

I also possess an M.Sc. from the Department of Computer Science at Bar-Ilan University, which was supervised by Prof. David Sarne of the Intelligent Information Agents Group (IIA).

Research interests:

AI-based behavior modeling
Human-Centered AI
Human-Computer Negotiation
Predicting online behavior using machine learning
AI-based user modeling
Game Theory
Fair Division
Churn Prediction

Research:

My research uses human-centered AI to extract the manner of human thought and the associated behavioral patterns applied in the solution of various problems and to develop interactive tools which assist humans in the completion of tasks based on that extracted information. A specific domain which I am interested in is the approaches and decisions of data scientists during the deelopment of predictive models and the development of tools which assist them in the development predictive models via the real-time prediction of moments in which they require assistance based on thier workflow.

Another area of research I am concerned with focuses on domains in which people are required to cooperate and negotiate with respect to the use of various resources with an emphasis on the development of real-world algorithms (e.g. profit-sharing in inter-company cooperation settings, room allocation and rent-sharing among housemates, the formation of government coalitions, etc.) that are concerned with the issue of the “human in the loop”.

More specifically, this research combines theoretical game theory solution concepts with machine learning techniques for predicting human behaviors. In addition, I develop intelligent agents that maximize revenues in various domains. Furthermore -- and importantly -- both are carried out with a particular emphasis on (human) user satisfaction with the algorithms' outcomes.

Teaching Experience:

I was a teaching assistant and professor in various computer science courses (see the list of courses in the teaching page)

Industrial Experience:

I presently work at Dicta as an AI researcher (see top of present page for details)

I worked as a Data Scientist at Diagnostic Robotics - a provider of an AI solutions in the healthcare domain that counts governments, healthcare service providers, insurance companies, and patients among its clients. More specifically, the company develops predictive models in various healthcare sub-domains such as differential ER (emergency room) diagnoses, hospitalization predictions, etc.

Another notable position I held was that of an artificial intelligence researcher at Amdocs, a leading developer of software for telecommunication providers with clients around the globe, where I developed a churn-prediction machine-learning algorithm which identifies and predicts at-risk maintenance contracts (contracts which are at risk of not being renewed by the customer).