Explore Alibaba’s New Open‑Source AI Framework, Massive Cluster Dataset, and Blink Engine

Alibaba Open‑Sources Its First Deep‑Learning Framework X‑DeepLearning

Deep learning has driven breakthroughs in speech recognition, computer vision, and natural language processing, powered by GPUs and open‑source frameworks. Existing frameworks like TensorFlow, PyTorch, and MxNet struggle with large‑scale industrial scenarios such as advertising, recommendation, and search. X‑DeepLearning is designed for these use cases, optimized through Alibaba’s advertising business, offering superior training scale, performance, horizontal scalability, and built‑in industrial algorithms for ad, recommendation, and search domains. The code is available on GitHub: https://github.com/alibaba/x-deeplearning .

Alibaba Releases a Real‑World Cluster Dataset (Alibaba Cluster Data V2018)

In 2015 Alibaba’s data center began co‑locating batch offline tasks and latency‑sensitive online services on the same machines, boosting overall utilization. After three years of testing and optimization, resource utilization rose from 10% to 45%, reducing average transaction cost during "Double‑11" by 17%. To enable researchers to study large‑scale data centers, Alibaba released the Alibaba Cluster Data V2018 dataset, detailing server and task information from a production cluster. Download link: https://github.com/alibaba/clusterdata/blob/v2018/cluster-trace-v2018/trace_2018.md .

Blink, Alibaba’s Internal Flink Fork, Will Be Open‑Source in January

Real‑time stream processing has become mainstream, with engines like Spark Streaming, Kafka Streaming, Beam, and Flink. Since 2015 Alibaba has been improving Flink and created an internal branch called Blink, serving core real‑time workloads such as search, recommendation, advertising, and Ant Financial. At the Flink Forward China summit on December 20, Alibaba announced that Blink will be open‑sourced in January 2019, aiming to deepen collaboration with the Flink community and promote adoption among Chinese enterprises.

FlyZoo Hotel – Alibaba’s First Future Hotel Opens in Hangzhou

The FlyZoo Hotel features full‑scene facial‑recognition check‑in, intelligent elevator control, robot bartenders, and robot room service. Guests can request items via Tmall Genie, with robots delivering them. The hotel operates without a traditional front desk, using service ambassadors and AI‑driven amenities, showcasing Alibaba’s AI and ecosystem capabilities.

A Letter to New Alibaba Recruits

As the 2019 campus recruitment season approaches, several Alibaba technology leaders—including the head of Alibaba Cloud AI, Ant Financial’s Vice CTO, DAMO Academy’s database chief scientist, and a senior researcher—have written a welcoming letter to new hires, sharing insights and encouragement for the upcoming journey.