Netflix By The Bay

Netflix’s presence at our conferences over the years has been both consistent and inspiring, providing updates on the latest developments and challenges in their machine learning and data engineering efforts.

RSVP

This year, Netflix is represented by its technology. Netflix has been a pioneer in open-source data infrastructure. They originally introduced Apache Iceberg, a high-performance table format for massive analytic datasets, which has since become a top-notch OSS project widely adopted across the industry. Iceberg enables structured, queryable datasets that power both large-scale analytics and modern AI applications.

This year Danica Fine and Josh Reini (Snowflake) will demonstrates how Iceberg can enhance retrieval-augmented generation systems by adding precise, structured intelligence to unstructured data. Their talk showcases the impact of Iceberg and Netflix’s early contributions to the project.

Most recently, in 2023, Jeremy Smith discussed the long-term evolution of software systems in Aging: Evolving Software, Tools, and Ourselves, making a compelling case for why maintenance—not just rewrites—matters deeply in sustaining engineering value over time.

Looking back at 2019, Jeremy Smith accompanied by Jonathan Indig addressed tooling gaps with Polynote, a polyglot notebook supporting reproducible work across Scala, Python, and SQL to boost developer productivity.

Earlier, in 2018, Julie Pitt reflected on the state of applied ML, likening its maturity to software engineering circa 1998, and raised the thought-provoking question: “How do we make humans BETTER through ML?” Her talk addressed operational pain points like collaboration, deployment, and trust.

Netflix’s engagement with the community goes back even further. In 2017, Roger Menezes unveiled Vegas, an open-source Scala library designed as the “missing MatPlotLib” for Spark/Scala, enabling engineers to better visualize and refine the recommendation models behind their global platform.

Across these years, Netflix’s talks have consistently highlighted the intersection of innovation, operational rigor, and human-centric thinking in machine learning and data engineering.

What’s Netflix focused on now?

Netflix’s journey is one of constant reinvention, and at By The Bay, we've had the privilege to witness—and learn from—that evolution in real time. Netflix doesn’t just apply machine learning to streaming recommendations—they use it to optimize promotional artwork, create personalized trailers, and even help inform which shows to greenlight. ML is not an isolated function at Netflix; it’s embedded in decision-making at all levels.

They are especially focused on:

• Foundation models: Netflix is exploring how large language models (LLMs) can enable new capabilities in search, support, and creative content production.

• Personalization at scale: Ongoing work in improving ranking, recommendations, and member satisfaction through better personalization.

• ML tooling: A continued investment in ML infrastructure like Metaflow, Polynote, and internal tools that abstract complexity for data scientists.

  
  

They’ve shifted from the idea of "ML as magic" to "ML as infrastructure"—standardized, tested, and reliable. They want models that scale across teams and deliver value consistently.

RSVP