January 2024 Summaries

Hex and Modelbit offer a streamlined approach to deploying machine learning models directly from data platforms, allowing data scientists to quickly build and deploy models with minimal code. In a demonstration, a simple classification model is created in Hex to predict NBA players' positions based on features like attempted three-point shots and scored points, using an XGBoost classifier that achieves a 62% accuracy rate. The model is then deployed to Modelbit with a straightforward Python function, enabling predictions to be made via a REST API or directly from Snowflake. This integration facilitates large-scale predictions without additional infrastructure, exemplifying how Hex and Modelbit simplify the process of building, deploying, and utilizing machine learning models in production environments efficiently.

StubHub's data team has undergone significant changes in the past year and a half, shifting from operational efficiency to a more strategic approach to data management. The company aims to balance speed and quality of data insights for its stakeholders throughout the organization. By building a continuous cycle of data analysis supported by Snowflake and Hex, StubHub's data team has increased both the speed and quality of insights. This transformation is unique but offers valuable lessons for other data teams in balancing operational efficiency with long-term initiatives and leveraging key tools to empower faster and higher-quality insights.

Hex Technologies has decided not to charge an additional fee for its AI-powered features, known as Magic, which enhance data analysis capabilities such as generating SQL, editing Python, and building charts. Initially uncertain about the costs and pricing for these features, Hex offered them for free during their beta phase and has now chosen to integrate them into the standard offering without requiring an "AI add-on." This decision stems from the belief that AI is becoming an integral part of software, much like cloud services did in the past, and suggests that AI functionalities should be considered a fundamental component of the product. Despite the potential revenue from AI add-ons, Hex prefers a straightforward pricing model, aiming to provide a premium product without confusing surcharges, as the cost of AI inference continues to decrease. Hex views AI as a primitive that enables various monetizable capabilities, but emphasizes that current AI features should not be treated as a monolithic add-on, choosing instead to offer them as part of their competitive seat pricing, which includes high-quality support and a commitment to simplicity.

To move a data team from reactive ticket-answering to strategic, decision-impacting work, focus on improving collaboration and communication with stakeholders. Encourage sharing of work through better processes and tools that facilitate feedback loops. Empower functional teams to explore their own intuitions and ask questions of the data by setting up "explore from here" types of playgrounds. Establish shared definitions for business concepts, using metric or semantic layers. Build a knowledge library for your organization by codifying insights gained by the data team. Finally, invest in the right tools that empower your data team to create, save, and share their work effectively.

Snowflake, a powerful SQL data warehouse, and Hex, a collaborative data workspace for Python and SQL, work well together due to several reasons. Firstly, Hex provides secure connection processes that allow users to query data without handling credentials. Secondly, it offers true first-class support for both SQL and Python, enabling seamless switching between languages within projects. Thirdly, Hex allows parameterization of SQL queries with Python variables, facilitating dynamic user input in queries and complex chained queries across multiple data sources. Fourthly, it powers complex analytical workloads while serving up data to end users as curated and interactive data apps, leveraging Snowflake's performance and scalability. Lastly, Hex supports Snowpark for Python API, enabling efficient execution of dataframes with distributed computing on the cloud data warehouse.