top of page

Manasi Vartak: Can AI Write Its Own Story? Unveiling the Power of Self-Documenting AI

Updated: Nov 12

Manasi Vartak is the founder and CEO of Verta, the Menlo Park, Calif.-based provider of the Verta Operational AI platform and Verta Model Catalog. Manasi invented experiment management and tracking while at MIT CSAIL when she created ModelDB, the first open-source model management system deployed at Fortune-500 companies and the progenitor of MLflow. After earning her Ph.D. from MIT, Manasi went on to data science positions at Twitter, where she worked on deep learning for content recommendation, and Google, where she worked on dynamic ad-targeting, before founding Verta.

Can AI Write Its Own Story? Unveiling the Power of Self-Documenting AI.

When working on a machine learning project, you'll likely use various vendors for tasks like data labeling, training and testing models, deploying them on different cloud platforms, and monitoring their performance. You'll also need to follow governance checks and handle documentation scattered across platforms like Confluence, Git and sporadically published papers. It's quite a beautiful mess!

How much time do you spend chasing down basic details about models, searching for documentation and dealing with the aftermath of inadequate records?

As the world embraces more ML and Generative AI-driven products, documenting AI systems becomes increasingly crucial. Model cards offer a great solution to the documentation nightmare, but properly documenting ML projects takes time and effort. The teams producing models are under such a time crunch, they rarely have any bandwidth left over to invest in model cards.

Given the efficacy of Foundational models in writing and summarizing text and code, in this talk we will showcase how we can use AI assistance to document AI projects. AI assistance can help by:

Self-documenting models from code, training data and other model attributes

Automatically capturing key model information with code summarization, model metadata, etc.

Preparing handoffs between ML, eng, product and other teams, with self-documenting model API contracts

Incorporating model documentation best practices with templates

bottom of page