Peng Xu

Retrieval-Augmented Generation (RAG) is a powerful AI framework that enhances large language models (LLMs) by integrating external context from retrieval systems, such as vector databases, into the generation process. This enables responses that are grounded, accurate, and context-aware. Rather than relying solely on pre-trained knowledge, RAG enables models to query their own knowledge base or document corpus. The retrieved context is then used to generate more relevant responses, complete with citations. In the following article, I explore a real-time RAG pipeline using a drag-and-drop interface using, 1. Google’s Agent Development Kit (ADK) - via ADK Web, CLI 2. Google Cloud Storage (GCS) - for document ingestion 3. Vertex AI RAG Engine - for chunking, embedding, and retrieval 4. Gemini 2.0 Flash LLM - for generating grounded responses This setup enables you to build a conversational agent capable of searching across multiple corpora (knowledge bases) and returning results that are relevant, factual, and properly cited. https://lnkd.in/e-FYz6Hp

23

1 Comment

🚀 Big news from Serve Robotics! 🚀 We’re thrilled to announce our new multi-year strategic partnership with DoorDash to expand autonomous sidewalk delivery across the U.S.! Starting in Los Angeles, DoorDash customers ordering from participating merchants may now have their orders delivered by a Serve robot 🤖 — with more cities coming soon. This partnership combines Serve’s reliable, sustainable autonomous delivery technology with DoorDash’s world-class logistics platform, creating a future where robots, Dashers, and drones work together to deliver faster, safer, and more efficiently. Serve has already completed 100,000+ deliveries from over 2,500 restaurants in cities including Los Angeles, Miami, Dallas, Chicago, and Atlanta — and this partnership marks another major milestone toward scaling that success nationwide. We’re incredibly proud of our team and partners who made this possible, and we can’t wait to continue redefining what’s possible in local delivery. 🌎💚 https://lnkd.in/gZkUu2RM #ServeRobotics #DoorDash #AutonomousDelivery #Partnership #Innovation #Robotics #Sustainability #FutureOfDelivery

66

How Uber Computes ETA at Half a Million Requests per Second Uber uses a graph representation of maps, where roads are edges and intersections are nodes. They partition the graph and pre-compute the best paths within each partition to handle a large number of ETA requests efficiently. Traffic information and real-time GPS data are also used to improve accuracy. Even a small improvement in ETA can lead to significant benefits for Uber, as millions of trips are made daily. You can read the full article in below link of system design newsletter. https://lnkd.in/dYRPJgyW

3

If your organisation is looking to leverage AI, they need the right tools to get AI Ready Data, and thats where DataOS by The Modern Data Company comes in.

2

The way live video gets delivered is about to change. Here’s what we’re building at Oracle. For 20 years, live streaming has run on the same fundamental architecture. It works — until it doesn’t. Peak events, thundering herds, last-mile congestion. Every major broadcaster knows the feeling when the big moment arrives and the cracks start to show. We’ve been working on something different. Dustin E. and I just published the first post in a series documenting Oracle’s foray into Media over QUIC (MoQ), a new transport protocol designed from the ground up for high-concurrency, real-time streaming. https://lnkd.in/gbg6vG98 #MediaTech #Streaming #LiveVideo #QUIC #CloudInfrastructure #OCI #VideoEngineering #MoQ

106

2 Comments

We co-authored this new white paper with JBF Consulting exploring what changes when transportation optimization moves beyond CPU-based architectures to GPU-accelerated optimization. One insight stands out clearly: the historic trade-off between optimization runtime, scale, and solution quality isn’t a law of physics - it’s an architectural limitation. In the paper, we go deeper into why CPU-based solvers hit that ceiling, how GPU-accelerated optimization changes the equation, and what that means for real-world, highly constrained transport networks. 👉 Read the full white paper - see the link in the first comment. #SupplyChainTechnology #TransportationOptimization #LogisticsStrategy #GPU #TMS #Chora #Tramm

10

3 Comments

"Vibe coding" transforms development, but can it handle enterprise rigor? I will be speaking at Google Cloud Next '26 alongside Google Cloud colleagues Oana Ionel and praveen killamsetti. Join us to explore how Gemini Cloud Assist and Antigravity agents bridge the gap in the world of AI Agents. We will demonstrate a complete lifecycle where Platform Engineers publish "Catalogs" of approved modules, allowing Developers to build intent-driven applications. You will see Cloud Assist analyze code to propose compliant infrastructure, automate IAM, remediate deployment failures, and generate CI/CD pipelines. This approach unifies velocity with governance, helping teams to stop fighting over tickets and start shipping with context. Explore the session library here: https://lnkd.in/dSzuNgHA.

148

7 Comments

In conversation with ITNews Asia, Oracle's Tirthankar Lahiri shares why 2026 could be a tipping point for intelligent data platforms. As enterprises operationalise AI, vectorised data, converged architectures, and built-in trust will define the next phase of innovation. https://lnkd.in/gNthADeV

1