DataDog's data scientist experiment and Postgres' forward march
As part of developing its new chat assistant tool, Bits, the company embedded a data scientists into the application team.
DataDog embeds its data scientists
For many years, data scientists and data analysts have been on a collision course in the work that they do within some companies.
The emergence of large language and diffusion models stands to shake things up a bit. And while responsibilities for managing and deploying those models could fall under a variety of roles—like machine learning engineering—we could start to see some changes in the ways companies manage their data science operations internally.
One example is DataDog, which is trying out a model where a data scientist is embedded within the team for a new chat assistant tool, Bits, which it announced this week at its conference in San Francisco.
Rather than relying on a sort of consultancy model—where a data science team handles inbound requests from other teams—they positioned a data scientist within the application team to manage evaluation, drift, and other parts of the product’s life cycle that are, well, data science problems.
“With recent work we have found it very, very useful to embed someone with the application team from the data science team,” Michael Gerstenhaber, VP of product at DataDog, told me. “They go to stand-ups and understand the concerns of the team, they help write the correct queries, and look at the same dashboards and own the outcomes of that product.”