Databricks

Coding tasks aren’t equally difficult, so why send every one to the same model? Smart Routing evaluates each task separately and selects the lowest-cost model capable of doing the job, helping balance model quality, latency and cost. In this demo, Omnigent splits an app build across planning, backend and frontend work, routes each task to a different model, and runs some of them in parallel.

Agentic apps are putting new pressure on the data stack. Join Databricks Co-founder and Chief Architect Reynold Xin to explore why Postgres is the foundation for a new data processing architecture, and how LTAP lets transactions and analytics read from the same copy of data. Learn how to: - Bring agent memory, application state and historical data under the same permissions - Remove sync intervals as a bottleneck in the agent memory loop - Maintain compatibility with existing Postgres drivers, extensions and ORMs You’ll also see the architecture in action through demos and enterprise patterns. Live November 10–12. Register now: https://lnkd.in/gvdwukWa

A big week for Databricks + Braze at #BrazeForge! Databricks VP & GM of CustomerLake Tasso Argyros joined Braze CTO Jonathan Hyman on the keynote stage to share how we’re connecting customer context, intelligence and action: - 𝗖𝘂𝘀𝘁𝗼𝗺𝗲𝗿𝗟𝗮𝗸𝗲: Bring customer context, profiles and audiences into Braze for orchestration - 𝗕𝗶𝗱𝗶𝗿𝗲𝗰𝘁𝗶𝗼𝗻𝗮𝗹 𝗢𝗽𝗲𝗻𝗦𝗵𝗮𝗿𝗶𝗻𝗴: Return engagement signals to Databricks without data duplication - 𝗕𝗿𝗮𝘇𝗲 𝗔𝗴𝗲𝗻𝘁 𝗖𝗼𝗻𝘀𝗼𝗹𝗲: Use Databricks Model Serving to give marketers access to every LLM when building customer journeys Together, we’re enabling end-to-end, 1:1 customer engagement that continuously learns and adapts. We’re also honored to receive Braze’s Technology Partner of the Year Torchie Award! 🏆

For decades, operational and analytical systems have been optimized separately for good reason. Transactions rely on fast row-based access, while analytics is optimized around columnar storage and broad scans. But now, AI agents are putting pressure on that boundary. They need to act on live operational data while also using data from the analytical side. LTAP changes where those workloads meet. It unifies them at the storage layer, with a hotter tier that keeps data in row format for operational access and a cooler tier that holds it in columnar format for analytical reads. Specialized compute can handle each workload independently. The architectural shift is simple: keep the specialized engine for each job while bringing the operational and analytical representations of the same data together underneath them. https://lnkd.in/gYgqMTgy

Two new frontier models just landed on Databricks. OpenAI GPT-6.1 Sol and SpaceXAI Grok 4.7 are both live today. GPT-6.1 Sol leads the cost-quality Pareto frontier on OfficeQA Pro v2. Grok 4.7 reaches the frontier at enterprise document parsing. Two benchmark leaders, available the day they ship. Frontier models are shipping faster than ever. Unity Gateway keeps you right there with them. It picks the right model for the right task, governs every call, and controls costs across GPT-6.1 Sol, Grok 4.7, and 60+ other frontier and open models already on Databricks. Try GPT-6.1 Sol and Grok 4.7 today: https://lnkd.in/gXu2JxKM

AI doesn’t have an intelligence problem. It has a context problem. When business context is scattered across dashboards, documents, tickets and chats, users struggle to get fast, trusted answers and data teams stay stuck fielding ad hoc requests. See what it takes to close that gap with a unified context layer that helps AI deliver trusted answers, act autonomously and give teams back time. Go under the hood of Genie Ontology and see the approach in action with a product demo. Watch on-demand: https://lnkd.in/gTkr_QN5

At Databricks, we want employees using the best models on Day 1. But new doesn’t always mean better, and rolling out a more expensive model across thousands of employees can quickly drive up costs. So when models like Opus 5.5 and GPT-6 Sol launch, we make them available quickly, evaluate them against real-world usage, and use Unity Gateway to manage access, spend and model selection at scale. Learn how the Databricks AI engineering team rolls out frontier models across the company and decides which ones belong in our AI stack: https://lnkd.in/gxSM2yD4

Anthropic's Claude Sonnet 5.5 is now available on Databricks across AWS, Azure and GCP, governed by Unity Gateway. It improves efficiency over Sonnet 5 for coding and agentic use cases, and reached Opus 5-level accuracy on document understanding, parsing and search. It joins Claude Opus 5.5, Claude Fable 5.1 and 60+ open-source and frontier models on Databricks. Build domain-specific agents with Agent Bricks, deploy them as Databricks Apps with Lakebase-powered memory, and govern every call through Unity Gateway. See documentation: https://lnkd.in/ePnCZe8A

We've invested a lot at Databricks to be able to roll new models out to every employee on day 1. Today we're sharing how we did it for the benefit of other companies. Why is it hard? Some new models are amazing (cheaper, better, etc) but some are actually worse! As an example: Opus 5.5 is a FANTASTIC model, but 5.0 was actually worse than 4.8 along most dimensions we measured (more expensive and less liked by users). Moving to bad models is a huge risk - you can blow up costs overnight or tank quality. But NOT migrating to great models is also a problem, you leave quality and money on the table... so how do you solve it?? Read our post! Bonus: Most of the techniques we've developed are also built into Unity Gateway! So you don't even need to read... just use the Gateway! https://lnkd.in/giMjNikp

AI agents need low-latency, high-accuracy search across massive datasets, and they can trigger thousands of concurrent retrieval requests in seconds. Traditional OLTP databases weren't built for that. Until now, solving it meant bolting a standalone search engine onto your primary database with an ETL pipeline. Lakebase Search is generally available. Two new Postgres extensions bring scalable vector and BM25 full-text search directly into Lakebase Postgres: - New frontier for price-performance, latency, and recall on VectorDBBench - 4x cheaper than running pgvector for the same workload - True pay-per-use with zero compute cost when idle https://lnkd.in/gHi7sbtM