Marco Rossi

Principal Data Scientist

关于

I’ve been in the Real-World Reinforcement Learning team for more than 6 years and we have productized some of the most research-oriented technology in the world.

I’m passionate about creating AI solutions for learning and decision making systems (from intelligent agents to A/B testing platforms). Over the years I’ve focused on many areas, like reinforcement learning, convex and global optimization, signal processing, and data science. I have a PhD in EE. A strong track record of project management, public speaking, mentoring, coaching, and academic publications. My papers have been citied more than 400 times, and I created many patents for Microsoft.

My mission is to empower people and organizations to elevate their decision-making. At MSR we have a world class team of researchers, PMs, data scientists, and engineers to advance AI in the real world. Part of my focus is to ensure that everyone in our team can thrive and provide their unique contribution to benefit Microsoft and to the world at large.

I played soccer at a professional level when I was in Italy and I love bringing my soccer ball everywhere I go (literally). Because of my connection with soccer, simply juggling the ball is an immediate flow trigger for me.

精选内容

Beyond A/B Testing: Offline Scorecards

A/B tests are the golden standard for production traffic owners (PM, stakeholders, website managers, etc) to find the best way to serve that traffic to improve a set of business KPIs – usually referred as scorecards. We propose a system that evaluates offline agents based on data collected by a Randomized Control Trial (e.g., an A/B test, Azure Personalizer, etc). The main goal of the proposed solution is to shorten the time necessary for production traffic owners to answer “what’s the best way to use my traffic to improve my scorecards” when compared with state-of-the-art A/B test technology.