Bitu Technology is the registered company of Tubi's China team. We are engineers passionate about leveraging advanced technology to improve lives, and we hope to use this channel to connect and advance together.
This article explains how Tubi customizes Spark SQL using lightweight macro‑based extensions to simplify column exclusion, JSON path queries, and other complex operations without modifying Spark's source code, detailing the two‑stage processing, example macros, and benefits for big‑data workloads.
The article details Tubi's ad‑serving pipeline from pacing control through bid collection, video transcoding, CPU‑intensive bidding optimization, result rendering, and final integration, highlighting the use of Akka, Spark, CRDTs, and gRPC in a distributed backend architecture.
The article shares Tubi's practical experience in building a fast‑iterating machine‑learning platform, emphasizing early measurement, simple end‑to‑end A/B testing, clear launch plans, lightweight popularity and embedding models, and rapid experimentation to drive product decisions.
The article reviews RecSys 2020’s shift to a virtual format, highlights the organizers’ use of tools like Whova and Gather.town, and summarizes several industrial and academic research breakthroughs presented at the conference, including PURS, behavior‑based popularity ranking, contextual item‑to‑item recommendation, counterfactual learning, debiasing techniques, and a large‑scale bandit dataset.
The article summarizes a series of technical talks held in August and September covering functional programming concepts such as Lens, Continuation‑Passing Style transformation, and F‑algebra with detailed explanations, demos, and theoretical insights for developers and researchers.
The article summarizes a 2020 Scala online meetup, covering Akka's distributed state‑management for ad‑serving, Parsec parser combinators for arithmetic expression evaluation, AST manipulation with ANTLR, and an introduction to the functional ZIO library, while reflecting on personal takeaways.
This article describes how Tubi’s data science team built a robust KPI forecasting system with Facebook’s Prophet, covering visualization dashboards, anomaly detection, feature engineering, PySpark deployment, and evaluation using Brier scores to improve business decision‑making.
Cat Talks is Tubi’s internal bi‑weekly series where engineers discuss foundational topics such as category theory, functors, natural transformations, free monads, parser combinators, and lenses, using live coding in Haskell and Scala to deepen understanding of functional programming concepts.
This article explains the challenges of debugging a wide range of OTT platforms, reviews existing tools like Eruda, vConsole, and ADB, and introduces a custom remote debugger built on Chrome DevTools Protocol that unifies debugging, JavaScript execution, and data inspection across devices.
This article details how Tubi built the Tubi Data Runtime platform on JupyterHub using Kubernetes, covering authentication with Okta SSO, custom Docker images, shared EFS storage, multi‑service support, GPU enablement, node affinity, cluster autoscaling, and monitoring with Prometheus.
