Auto-Tuning of Copilot Prompts
AI-agent-based tool for tuning and optimizing system prompts, and particularly those of Microsoft Copilot offerings. 新しいタブで開く
Meeting Intelligence: Task Rephrasing
As part of the general theme of Meeting Intelligence, we extract and rewrite tasks and to-do items identified in meeting transcriptions. 新しいタブで開く
mai: MAth Interpreter
Educational tool for computer-assisted mathematics, wherein the language of mathematics is defined from first principles. Using `horc’, a fully operational proof assistant is obtained at no additional cost.…
horc: HORn-Clause Interpreter
Implementation of `pure’ logic programming as a lightweight extension to Prolog. It is a minimalist (meta-)language allowing to jointly define and implement more complex formalisms.
概要
I am a Principal Researcher at Microsoft, specializing in generative AI and machine learning. I bring nearly a decade of experience in generative ML, and more than 18 years across industry and academic research.
At Microsoft, and previously at IBM Research, I have led research initiatives and teams across machine learning, NLP, and AI. My work spans science, technology, and product — from research published in leading academic venues to large-scale AI systems and agents used by millions of people worldwide, including ongoing contributions to Microsoft 365 Copilot. I have worked hands-on across several generations of ML technology, from classical machine learning and LSTMs to modern LLMs and AI agents.
My background is in mathematics and computer science. I hold a Ph.D. from the Weizmann Institute of Science and a B.A. from the Technion, with earlier research in logic and formal methods.
I live in Haifa, Israel. Outside of work, I enjoy strength training and following recent developments in digital mathematics (opens in new tab).
You can also find me on LinkedIn (opens in new tab), GitHub, (opens in new tab) and Google Scholar (opens in new tab).