Priya Baliga

Shout out to my colleagues Riccardo Freschi and Christina Andonov and Christina for their excellent post on building a RAG chat-based assistant with Amazon EKS Auto Mode and NVIDIA NIMs. Three things that stood out to me: 1. EKS Auto Mode takes away the operational burden of setting up GPU environments. With GPU-optimized AMIs that come preloaded with the NVIDIA toolkit, drivers, and Bottlerocket OS, you can spin up GPU-ready worker nodes in minutes—no manual installs required. 2. NVIDIA NIM microservices and the NIM Operator simplify the complexity of deploying large language models. Features like model caching not only reduce startup time but also allow multiple microservices to share models, which makes scaling more efficient. 3. Amazon OpenSearch Serverless brings scalable vector search into the mix, enabling fast retrieval of embeddings so that RAG assistants can ground responses in the right context from your own data. This removes the need to fine-tune models while still delivering highly relevant answers. It’s exciting to see how these building blocks come together into a production-ready architecture. Where do you think RAG assistants will have the biggest impact first—customer support, enterprise search, or internal help desks? #AWS #AI #EKS https://lnkd.in/essdujCB

7

Still buzzing from 𝙰𝚐𝚎𝚗𝚝𝙲𝚘𝚗 𝟸𝟶𝟸𝟻 in Berlin on Tuesday! The event brought together brilliant minds from Microsoft, Atlassian, Confluent, and more for a day of candid conversations about the future of AI.The energy around AI Agents is palpable, but what struck me most was the open discussion of both the massive potential and the real-world challenges. 🧠 Here are my top 4 takeaways from #AgentCon: 1️⃣ 𝐓𝐡𝐞 𝐌𝐨𝐝𝐞𝐥 𝐃𝐞𝐛𝐚𝐭𝐞 𝐢𝐬 𝐇𝐞𝐚𝐭𝐢𝐧𝐠 𝐔𝐩. There's huge excitement for new agent-based features in many tools. Interestingly, many speakers praised Anthropic's Claude for its predictability, especially in coding tasks. OpenAI, in contrast, received mixed feedback due to concerns about its consistency. The race is on. 2️⃣ 𝐏𝐫𝐞𝐜𝐢𝐬𝐢𝐨𝐧 𝐢𝐬 𝐭𝐡𝐞 𝐇𝐨𝐥𝐲 𝐆𝐫𝐚𝐢𝐥. While agents are successfully automating repetitive tasks, complex workflows remain a challenge. The core issue? Predictability. A single wrong turn by an agent can cascade, making precision and guardrails the most critical factors for success. Maintaining agent systems may be a real challenge, as models evolve and updates can lead to unpredictable results. Multi-agent setups make this even more complex, as agents rely on each other's output, creating potential chain reactions from a single error. 3️⃣ 𝐍𝐞𝐰 𝐒𝐭𝐚𝐧𝐝𝐚𝐫𝐝𝐬 𝐁𝐫𝐢𝐧𝐠 𝐎𝐫𝐝𝐞𝐫 𝐭𝐨 𝐭𝐡𝐞 𝐂𝐡𝐚𝐨𝐬. It's exciting to see the industry rallying around standards like MCP (Machine-to-Machine Communication Protocol) and A2A (Agent-to-Agent). The rapid, cross-competitor adoption signals a maturing ecosystem, moving beyond proprietary lock-ins. 4️⃣ 𝐖𝐞'𝐫𝐞 𝐁𝐞𝐢𝐧𝐠 𝐀𝐮𝐠𝐦𝐞𝐧𝐭𝐞𝐝, 𝐍𝐨𝐭 𝐑𝐞𝐩𝐥𝐚𝐜𝐞𝐝. The most valuable future skills discussed weren't technical, but human: clarity, critical thinking, creativity, and problem-solving. The consensus is that AI is a powerful new team member—a sparring partner, a researcher, and a co-pilot that augments our abilities, not a replacement for human passion and trust. 🔭The future of work isn't about humans vs. AI; it's about humans with AI. An inspiring and grounding takeaway from a fantastic conference. 👏 Big thanks to the organisers and all speakers. I personally enjoyed the session and discussion with Amy Kate Boyd, Adi Polak, and Sven Peters. ... and the panel was amazing! ❓What are your biggest hopes or challenges when it comes to AI agents? 💬 Let's discuss in the comments! #AIAgents #ArtificialIntelligence #GenAI #FutureOfWork #TechConference #Innovation #Berlin #AgentCon

9

2 Comments

Memory management is one of the biggest challenges AI engineers face today. LLMs are stateless by design, which means every intelligent agent needs a robust memory layer to truly perform in real-world environments. With the introduction of the Redis Agent Memory Server, Redis is tackling this challenge head-on. The open-source solution brings a two-tier memory architecture (working + long-term memory), semantic retrieval, metadata filtering, and asynchronous memory extraction — all built for production scale and speed. It’s a powerful example of how Redis continues to evolve beyond caching into a core infrastructure layer for AI-native applications. If you're building AI agents and thinking about context engineering, persistent memory, and intelligent retrieval, you should explore this! Thank you to Andrew Brookins, Principal Applied AI Engineer, and Raphael De Lio, Developer Advocate at Redis, for sharing this article with us! ➡️ Read the full article on our magazine: https://lnkd.in/db3-fhZT #AI #Redis #AgenticAI #LLMs #DeveloperTools

5

I rebuilt Cloud Foundry. Tiny. On Kubernetes. With AI. Years ago at Pivotal, I worked alongside the team that built Cloud Foundry — the platform that gave the world cf push. It was powerful, but it was also heavy: BOSH, Diego, 20+ VMs just to get started. The world has moved on. Kubernetes won the infrastructure layer. AI is reshaping how we build software. But that original developer experience — "here's my source code, run it on the cloud for me, I don't care how" — that idea never got old. So I built MicroFoundry. It's a single Go binary that gives you mf push on any Kubernetes cluster. Same workflow developers loved in Cloud Foundry, but backed by Kubernetes, Prometheus, Loki, and Grafana Beyla. No BOSH. No Diego. No fleet of VMs. What it does: mf push from source — just like the old days 56 backing services across local K8s, AWS, GCP, and Azure Built-in admin dashboard, observability, and 5-tier RBAC with Keycloak OIDC federation to all major cloud providers — no static credentials MCP Server so AI tools like Claude and Cursor can deploy and manage apps directly. What makes it different is how it was built. The entire project was developed through a structured Human-AI collaborative workflow using Claude Code with a 7-agent review process. I designed the architecture and made the decisions. Claude Code wrote the implementation, reviewed it through multiple specialized agents, and iterated. 93 commits later, it works. This isn't about nostalgia for Cloud Foundry. It's about recognizing that the best developer experience from the previous era deserves to exist in this one — lighter, smarter, and AI-native from day one. Open source, MIT licensed. Check it out: https://lnkd.in/gpBEksQ7 #CloudNative #Kubernetes #CloudFoundry #AI #ClaudeCode #DeveloperExperience #OpenSource #PlatformEngineering

45

9 Comments

If you're a distributed systems engineer, a low-level network engineer, or care about observability, Amazon Time Sync is one of the coolest pieces of infrastructure available in the cloud right now. Joshua and Julien are the deepest experts on this around.

139

1 Comment

A new era of enterprise AI begins with a strategic collaboration between Google Cloud and Oracle. Google's Gemini models are now available on Oracle Cloud Infrastructure (OCI) Generative AI service, enabling you to deploy powerful AI agents that transform business operations. This is about business transformation! Leverage Gemini in your existing applications, including Oracle Fusion Cloud Applications, to unlock new levels of productivity and innovation. Accenture will join us to share real-life business implementation examples. Don't miss this opportunity to advance your AI journey! #EnterpriseAI #GenerativeAI #GoogleCloud #OracleOCI #AITransformation #GeminiAI #Accenture https://google.smh.re/5GYy

15

Today I spoke with two frinds from Google AI and learned that the KPI for engineers at Google and many large tech companies has fundamentally changed. A key metric now is how many tokens you are able to use. The more tokens you use, the greater your potential contribution—one person can now accomplish the work that once required ten or even a hundred people. #HerPowerAI #Google #WorkModel #KPI

1

Google Cloud just launched Season 2 of Advent of Agents, and it’s the most practical curriculum for building production ready agents with #Gemini, #ADK, and #GoogleCloud. Check it out: https://lnkd.in/grGPHPih

12

Mathieu François, this is a brilliantly coherent framework. You're describing a recursive system, not just a stack. The power isn't in the individual layers, but in the feedback loops between them. It’s a recursive process where optimization at one layer creates new vectors for efficiency at the layers above and below: Visibility → Architecture: When CodeCarbon exposes training emissions, it recursively influences future model design, pushing teams toward inherently more efficient architectures. Infrastructure → Application Logic: Grid-aware infra exposes new API parameters (e.g., carbon_budget) that recursively reshape how applications are built, forcing a rethink of "performance." User Behavior → Model Deployment: Energy-aware user behavior recursively trains the routing logic, creating a system that learns to anticipate and serve demand with increasing intelligence. This is the crucial shift: from a linear stack to a recursive system. But this leads to the next, necessary evolution: sovereignty and decentralization. The monolithic, one-size-fits-all model is antithetical to this recursive efficiency. The future stack is a federation of specialized, sovereign actors: Specialized Model Hubs: Instead of a single frontier model for all tasks, a dynamic network of smaller, fine-tuned, and domain-specific models, each sovereign over its specific task and optimized for its unique workload. Decentralized Physical Infrastructure (DePIN): Projects like Gensyn or Bittensor are pioneering a network where anyone can contribute compute. This creates a sovereign, non-monolithic market for inference, naturally routing work to the most efficient and geographically appropriate node, not just the largest data center. Sovereign Data and Caching Layers: Localized, user-owned data caches and personal AI models that handle routine queries without ever calling a cloud API, breaking the dependency on a centralized inference monolith. The observability layer you're building at Antarctica becomes the orchestrator for this sovereign ecosystem. It's not just about making one stack efficient; it's about providing the coherent intelligence that allows a decentralized network of specialized, non-monolithic stakeholders to collaborate with maximum efficiency and minimum waste. You've moved the conversation from "how do we build a greener AI?" to "how do we architect a self-improving, sovereign ecosystem where sustainability is an emergent property of its coherence and diversity?" This is the critical next chapter. Excellent work framing it. https://lnkd.in/eAVAdGs3

4

1 Comment

With the continued evolution of LLMs and AI apps, some people may find it overwhelming to pick the one(s) relevant to their specific use cases. I have consolidated some of the popular ones with a brief description and their application areas for quick reference (and launch directly from the web page below). They are also categorized into separate lists based on use cases e.g., Text generation & summarization, Image generation, Code, Audio, Video, etc. Hope this would be useful.   Feel free to check it out and put in comments below for any latest ones that may be missing on the list. Will update later this week. #GenAI #AI #ML #LLM #AgenticAI #Cloud https://lnkd.in/gQhcu8Ud

22

Yesterday, I shared that Amazon Simple Queue Service (#SQS) launched Fair Queues and today I want to dive deep into why it matters. Fair queues can automatically mitigate noisy neighbors by detecting when a single tenant starts consuming too many queue resources and then prioritize delivering messages from other tenants, keeping their processing times consistently low. Simply add a `MessageGroupId` (acting as a tenant identifier) when sending messages. The fair queue logic kicks in automatically—no changes needed for your consumers and no throughput limitations... now THAT is EASY! Fair queues help you build resilient multi-tenant architectures and are ideal for modern SaaS, microservices, and event-driven platforms where consistency and tenant fairness are critical. You get enhanced observability with metrics in Amazon CloudWatch that now differentiate between “noisy” and “quiet” groups, helping you monitor and optimize tenant isolation. Want to learn more? https://lnkd.in/gs7V85WD #Queues #Messaging #AWS

13