Automating Investigation at Datadog

As a Staff Product Manager of Investigation and Search at Datadog, I work on enabling users to find answers in complex systems.

My initial focus was Data Visualization and Querying, where I first focused on creating more powerful and expressive data analysis tools. In parallel, I studied how customers move through an investigation. As my understanding in this space matured in my first year at Datadog, I developed a vision for how to automate investigation.

In year two, I conceived and shipped Datadog’s first widely used AI feature, Watchdog Explains. In near-realtime, this feature automatically detects anomalies on dashboards, and picks apart the exact dimension contributing to the anomaly. The feature was a good start for automating investigation: it's efficient enough to run for free, and shows few false positives.

My focus now is moving beyond triage to automating root cause analysis. I’m working through challenges like: how do we proactively scan for issues without incurring too much cost at Datadog’s scale? How do we take advantage of an LLM’s creativity while ensuring it returns quick and accurate results? And how do we bring in the user's expertise without making them do too much work?

My team has expanded to include Search, in order to build context layers and semantic search models to ground an LLM in its investigation. We are working on building a multi-layered investigation system that takes multiple, gradually more comprehensive passes at detecting an issue. The user is still central to the investigation, akin to a text-based adventure game.

Teaching students to use code as a creative medium

I teach Creative Computing at CUNY Graduate Center, where grad students learn to code as a medium for creative expression. We explore how computation can make things that couldn't exist any other way. I love seeing students discover what becomes possible with code.

At Parsons, I got to explore interactive art

After years of designing and building interactive apps on the side, I went to grad school for Data Visualization to scratch my creative itch.

My thesis at Parsons, How We Gaze, was a gallery of gazes: it distorts art to reflect how the viewer looks at it in realtime. How We Gaze was selected as a winner for the Pudding Cup, an independent data visualization award. Snippets of this project and others below.

I launched two healthcare products at Arcadia from the ground up

I started my career at Arcadia, a healthcare tech startup. I got my hands dirty, teaching myself to design, code, model data, and pitch new product ideas. My obsession with making sense of the data we had led me to create the visuals below.

At Arcadia, I launched two enterprise workflow products from 0 to $XM: Care Manager helps care teams efficiently manage large patient caseloads with predictive risk models and automated workflows, and Referrals streamlines specialist referrals by automatically suggesting high-quality providers based on patient needs (see press).