Streaming entertainment, games and live programming
Netflix
Netflix offers an entertainment service where members can watch TV series, films and live programming, and play games across a variety of genres and languages. Its business also includes original content production and advertising. Original series became part of its offering in 2013, with titles including House of Cards and Orange Is the New Black.
Founded in 1997 by Reed Hastings and Marc Randolph, its first CEO, Netflix opened its online DVD rental store in 1998. Customers ordered discs for delivery and returned them in prepaid mailers. Personalization was already part of that service: registered shoppers received recommendations based on previous shopping sessions. Today, co-CEOs Ted Sarandos and Greg Peters combine backgrounds in content leadership and product, technology and operations.
Netflix reported more than 325 million paid memberships at year-end 2025 and approximately $45.2 billion in annual revenue. Advertising revenue exceeded $1.5 billion. Its agreement to acquire Warner Bros. Discovery’s streaming and studios businesses was terminated on February 27, 2026; Paramount Skydance paid Netflix a $2.8 billion termination fee on Warner Bros. Discovery’s behalf.
3 talks
Newest first
3 speakers at AIE
Affiliations reflect their AIE appearances, not necessarily current employment.
Start here
- The Infinite Software Crisis
Start with the essay based on Jake Nations's talk to distinguish structural simplicity from ease of implementation and separate essential requirements from accumulated complexity.
Jake NationsAI Engineer Code 2025
- AI Agents for Performance: Ship Faster, Pay Less
See how a quadratic-time code pattern surfaced in profiling and how optimization was assessed through measured production savings.
Rajat ShahAI Engineer World's Fair 2026
- One model to rule recommendations: Netflix's Big Bet
Learn how tokenization, multi-token prediction, and generative retrieval fit into a shared recommendation foundation model.
Yesu FengAI Engineer World's Fair 2025
Messages from the stage
Understanding before automation
Nations advocates separate research, planning, and implementation phases with human validation. His authorization migration example shows why manually uncovering hidden constraints can be necessary before automating a change.
Turn profiling discoveries into reusable knowledge
Shah connects CPU profiles and call stacks to the exact deployed Git revision. A shared anti-pattern catalog carries performance findings into coding and review across services.
Shared representations for personalization
Feng describes autoregressive transformer training on multifield user-interaction events. Combining ID and semantic embeddings addresses cold starts, while shared representations support varied recommendation requirements.
Affiliations reflect each recorded session, not necessarily current employment.