Sanjana Gautam

I am a Senior Applied Scientist at Microsoft, on the ERP AI team building the Finance Agent, an AI agent for finance professionals. My work centers on evaluation: I design scientifically grounded methods to measure and improve how these agents perform, using LLM-as-a-judge frameworks, automated experimentation, and systematic prompt, data, and instruction tuning.

In my work, I approach applied AI through a human-centered research lens. My background blends human-computer interaction (HCI), AI ethics, and behavioral science. I care about building AI systems that are not only capable but trustworthy, interpretable, and genuinely useful in real workflows. I am especially interested in how automation reshapes knowledge work, how AI tools change the way people reason and decide, and how we design them to balance efficiency with rigor and integrity. I also remain drawn to AI’s role in broadening access to knowledge through open-source tools, data sharing, and educational initiatives.

Before joining Microsoft, I was a Bullard Research Fellow at the University of Texas at Austin’s iSchool, where I collaborated with Dr. Matthew Lease and contributed to the mission of Good Systems, examining the unintended consequences of AI deployment. I was also engaged with the Cosmic AI Institute, exploring how trustworthy, interpretable AI methods can support scientific discovery in astronomy. I completed my PhD at Penn State’s College of Information Sciences and Technology under the supervision of Dr. Mary Beth Rosson, as an active member of the Innovation and Collaboration Lab led by Dr. Jack M. Carroll and Dr. Rosson.