Bruno Aziza

For AI to truly scale in the enterprise, it needs more than just "wow factor" — it needs software engineering rigor. Snowflake CEO Sridhar Ramaswamy highlights that unlike consumer apps, enterprise AI requires a higher bar for accuracy and reliability. By applying the same discipline to AI that we apply to core software engineering, organizations can build systems that are resilient, documented, and trusted. Check out the full interview for insights on balancing experimentation with discipline. ⬇️

5

Awesome post by the Databricks team https://lnkd.in/gGQ9ptRe My summary: They trained a model called KARL that beats Claude 4.6 and GPT 5.2 on enterprise knowledge tasks (searching docs, cross-referencing info, answering questions over internal data), at ~33% lower cost and ~47% lower latency. The key insight: instead of throwing expensive frontier models at enterprise search, you can use reinforcement learning on synthetic data to train a smaller model that's faster, cheaper, AND better at the specific task. RL went beyond making the model more accurate. I t learned to search more efficiently (fewer wasted queries, better knowing when to stop searching and commit to an answer). They're opening this RL pipeline to Databricks customers so they can build their own custom RL-optimized agents for high-volume workloads. I think we'll continue to see data platforms become agent platforms. Databricks' KARL paper is really an agent platform play. The pitch: you already store your enterprise data in the Lakehouse, now Databricks will train a custom RL agent that searches and reasons over it, tuned specifically for your highest-volume workloads (workloads = apps = agents). The business move is closing the loop: data storage → retrieval → custom agent training → serving, all on Databricks. They're turning "your data lives here" into "your agents live here too." Kudos Ali Ghodsi Reynold Xin Matei Zaharia

56

1 Comment

.@Snowflake delivers strong Q4 amid data, AI demand https://bit.ly/4cPuNeX Snowflake reported a better-than-expected fourth quarter and said it added 740 net new customers, up 40% from a year ago. @ldignan

10

Quick little demo using IBM watsonx orchestrate to navigate Confluent Cloud via MCP. This demo uses Confluent's local MCP server with tools to explore, manage, and debug real-time data streams. You can find more about the MCP server here: https://lnkd.in/guPksxN6 This is just the start, lots more to come, bringing watsonx agents and real-time together.

161

2 Comments

It is clear that strategic AI enabling offerings are gaining traction. In Azure, demand continues to outstrip supply, particularly for key data and AI products such as Fabric, Foundry, Copilot Studio and Agent Builder. Without supply chain constraints, Azure would have grown 40%+. This presents a higher likelihood that Azure can achieve that growth threshold near term. In the office business, commentary suggests that Copilot is ramping nicely, with 4.7 million subscribers, +75% y/y. While the outlook for capex to decline sequentially in Q3 is unusual, the shift to short lived assets (such as GPUs and CPUs) validates that supply chain constraints could be reversable in the near term.

2

Oracle’s GoldenGate Stream Analytics 26ai accelerates real-time intelligence at scale - and Goldenore ensures the architecture, integration, and execution, align with enterprise performance, resilience, and scale. Happy to share how Goldenore has deployed OGGSA.

9

🔌 Another signal that the AI compute race is rewriting the semiconductor and infrastructure landscape. Bitcoin miners have already signed more than $65 billion in AI data center contracts, and we are increasingly seeing the spillover into the broader Bitcoin mining ecosystem. Auradine’s announcement that it is evolving into Velaura AI is another clear signal of where the industry is heading. The company built expertise around energy-efficient ASIC design, proven through millions of chips deployed in Bitcoin mining hardware. That engineering DNA is redirected toward AI compute. In 2025 Auradine raised $153M in Series C funding and simultaneously launched AuraLinks AI, a business unit dedicated to AI data-center networking. The new branding signals a strategic emphasis shift, where AI becomes the primary narrative instead of mining. 💡 In many ways, mining became an unintentional proving ground for energy-efficient compute: 🔸 Designing chips that deliver maximum compute per watt 🔸 Operating infrastructure at extreme power densities 🔸 Optimizing hardware economics in energy-constrained environments Those capabilities now sit at the center of the AI scaling challenge. The line between “mining infrastructure” and “AI infrastructure” is becoming increasingly blurred. We are already seeing: ⚡ Mining operators exploring AI and HPC workloads ⚡ Container and infrastructure providers building multi-compute data centers ⚡ Management software platforms evolving to support multiple compute workloads ⚡ Hardware brokers increasingly operating across both ASIC and GPU markets Auradine’s evolution into Velaura AI is another step in that broader trend. The future of compute will not just be defined by faster models, but by how efficiently those models can run at scale. And the companies that spent the past decade solving compute-per-watt challenges may be better positioned for the AI era than many people expect.

4

Generative AI only works if the data behind it is trusted. We partnered with Elastic to help teams ground AI answers in real enterprise data. Dogukan Yazici, Partner Sales Engineer at Alteryx , breaks down how it works and why it matters. Read Here: https://ow.ly/k9o530sSa5H #AI #Elastic #Alteryx

8