Chang(chauncey) Liu

Interesting article from Bloomberg about life at Google's X in a time of layoffs and cost reduction at https://lnkd.in/gmw3GG7C. X is a component of Alphabet (Google's parent company) that incubates new ideas. They have been responsible for some great successes - getting Waymo and a lot of Google's AI work started - but they've also had projects that didn't survive - like Google's attempt to provide Internet access via weather balloons. When you work at a company that has hundreds of billions of dollars in revenues, it's hard to get a lot of interest in an entirely new business idea that brings in hundreds of millions. If you have a project that improves ads revenue by that amount, or decreases capital expenditures, you can do it, but if you want to, say, build wearable tech, it becomes hard to attract the attention of marketing or product. So, new ideas often die on the vine. For the most part, this is good and proper. Companies may have the financial wherewithal to branch out, but more often than not, they consider it a distraction. And they are probably right - it's hard to argue that Meta was right to divert such a huge chunk of their resources to chase after the metaverse. They've (rightfully) pulled back, and are more focused on their core business again. On the flip side, if you are in the tech sector, and you don't innovate, you are likely to get left behind at some point. So, you need to make and manage these investments somehow. When I joined Google, the CEO had a 70-20-10 rule - 70% of effort went to core businesses, 20% to broader efforts to improve Google, and 10% to stuff that's completely blue-sky. This is probably not completely right, but it is also probably not completely wrong. X is the division of Alphabet that is *specifically* designed to bring blue-sky products to life. Unsurprisingly, they've had to pivot and rethink their strategy many, many times. The article interviews a number of people who have worked at X over the years; it's well worth a perusal, for folks who are interested in the story.

18

2 Comments

Our SIGMOD '24 paper explores MemoryDB's architecture, which delivers in-memory performance and strong consistency at scale. A key innovation is the decoupling of durability and replication from the in-memory database engine, which allows scaling compute and storage independently based on customer workload needs. Customer data integrity, security, and availability are paramount for MemoryDB. The paper dives deep into innovations that the team built to make that possible (a personal favorite is off-box snapshotting which allows us to create consistent database snapshots on separate clusters to avoid impacting live customer traffic and preventing resource contention). We also share hardened operational experiences from running a distributed database at cloud scale. #aws #memorydb #inmemory #databases #durability #sigmod

91

1 Comment

Want to learn how to customize GPU memory management in Python and C++ code? Check out this #GTC24 session for hints and tips to help you handle large data more efficiently. Watch now.

3

Can you believe it's already been two weeks since the the first-ever #KafkaSummit Bangalore? ICYMI we have the full rundown of all the key moments from the premiere event for the #ApacheKafka community. We’ve got: 🔑 Keynote recaps (Big product announcements!) 💡 Session highlights (Flink, GenAI, and of course Kafka are top of mind) 🚀 Women in Technology spotlight ➕ And MORE Find it on our blog!

2

Check out the "Titan Text Premier Demo" built by my team on this page

9

Today, Amazon is announcing the Amazon Trusted AI Challenge, a global university competition to drive secure innovation in generative AI technology. This year’s challenge focuses on responsible AI and specifically on large language model (LLM) coding security. @Amazon @AmazonScience #Amazon #AIChallenge #LLM #GenerativeAI #GenAI #ResponsibleAI

1

Discover the optimal methods for deploying LoRA-tuned adapters with the Llama 3 NIM to maximize performance across various languages. Read more for developer insights > https://nvda.ws/4bGc01j

2

This year’s UC Berkeley AI Hackathon has concluded, and there are plenty of great moments and experiences to remember. Let’s take a closer look at the innovative ways three teams used Intel Tiber Developer Cloud. 👉 Dispatch AI: Reimagining emergency response with an empathetic AI-powered system. Dispatch AI leverages advanced technologies to enhance the 911 call experience, providing intelligent, emotion-aware assistance to both callers and dispatchers. 👉 Batteries by LLM: Employing LLMs to predict the structure of new lithium-ion electrolytes from text description and optimizing it further using first principles modeling to fight climate change. 👉 Accel: Deconstructing complex chemistry topics into easy-to-understand segments, ensuring that students grasp foundational concepts thoroughly. Congrats to the winning teams, and a big thank you to the more than 1,300 students who attended this year’s hackathon! Read the blog and catch up on all the best moments from the event. https://intel.ly/4cZjduG #HackWithIntel #ArtificialIntelligence #IntelDeveloperCloud

23

🚀 Exciting News from Uber Engineering! 🚀 We're thrilled to announce our partnership with Google Cloud Platform to modernize Uber's batch data infrastructure. As one of the largest Hadoop installations globally, hosting over 1 exabyte of data, our goal is to enhance our big data capabilities and keep pace with Uber's growing demands. 🔹 Migration Strategy: Our initial migration focuses on leveraging GCP's robust IaaS for our data lake, ensuring minimal disruption and maintaining our commitment to efficiency and security. 🔹 Future Plans: This move will not only boost our engineering velocity but also improve cost efficiency and expand data governance. We're setting the stage for a series of innovations and greater productivity across teams. Stay tuned as we share our progress and insights in a series of upcoming blog posts. Here's to making data-driven decisions faster and more accessible than ever! 🌐 #UberEngineering #CloudComputing #DataScience #GoogleCloud #BigData

122

4 Comments

Discover how #NVIDIAInception startup LiveX AI Inc. leveraged NVIDIA AI and GKE infrastructure to build #generativeAI agents that provide real-time, human-like customer experiences. 💡 Using NVIDIA A100 GPUs and NVIDIA NIM microservices, LiveX AI achieved a 6.1x acceleration in response generation speed for their AI agents.

4

1 Comment

I produced a demonstration video showcasing a personal assistant implemented by me on an AI terminal device, such as an AI-enabled PC. In this video, I illustrate how leveraging Nvidia technology and the LangChain framework enables efficient utilization of large language models within our application. By utilizing Nvidia-accelerated data centers, we call open-source models, including Mistral, to process user queries through retrieval-augmented generation. This approach effectively organizes and synthesizes information from the web, delivering users timely and accurate responses. demo video: https://lnkd.in/gUtxKBHf Step-by-step tutorial: https://lnkd.in/gcv8NBNM source code: https://lnkd.in/gPw8XjMj #NVIDIADevContest and #LangChain NVIDIA AI @NVIDIAAI NVIDIA Taiwan Happy AI hacking!

5

1 Comment

Our Kumo.AI blog post on Hybrid GNNs is out. https://lnkd.in/gpzr8TiF We have seen success with our Hybrid GNNs across several marketplace problems. Hybrid GNNs are a novel innovation at Kumo.AI because they model both repeated and explorative user interactions within a single GNN framework. The GNN automatically infers whether a user tends to do repeat purchases or tends to be more exploratory (ie buying new unseen items) and learns a repetition scalar per user. What we love about this approach is that it takes away from the standard recommendation systems architecture where you have multiple candidate generators - some for new unseen items, some for repeat behavior and then have to combine them in a final multi-objective optimization framework. Such systems which are a combination of multiple systems are hard to maintain, re-train and operationalize. On the Kaggle H&M challenge where most winning teams usually have approaches that require large amounts of feature engineering and the final system is a complex ensemble of approaches needing months to build, the Kumo.AI approach which is a 3 line predictive query that runs a Hybrid GNN under the hood is top 1%. Thanks Matthias Fey and Weihua Hu for the algorithms and innovation.

156

1 Comment