Ali Ghodsi

As more of you leverage foundational models in your organizations, remember they weren't trained or tuned on your internal metadata and taxonomies. Get a headstart with the metadata that naturally sits inside of Astronomer and is kept up-to-date automatically as your organization evolves. Companies leveraging orchestration metadata and lineage achieve higher quality answers with a fraction of the tokens. More to come!

9

Incredible conversation with two leaders in the AI space. Konstantine Buhler and Kari Ann Briski talk about the challenges and opportunities in the AI ecosystem, how open-source is changing the game, and what they see next (hint, it's around world models). Be sure to check it out. #opensourceAI #agenticAI

14

✅ Day 138 of #gfg160 Challenge 🌟 Problem: Perform Depth First Search (DFS) traversal of a connected undirected graph given as an adjacency list. We start from vertex 0 and must visit vertices from left to right exactly as given in the adjacency list. 🔍 Understanding the Task: We are given: A 2D adjacency list, where adj[i] contains all the vertices connected to vertex i. A connected graph — so all nodes are reachable from the starting node. We must start DFS from node 0. The DFS order depends heavily on the adjacency list order, so we must process neighbors in the given sequence without rearranging. 🛠 Logic: Track visited nodes → Use a list visited of size V to avoid re-visiting nodes. Recursive DFS: Visit the current node → mark it visited → add it to the result list. Recursively visit all unvisited neighbors in order. Start DFS from vertex 0. 📌 Example: Input: adj = [[2, 3, 1], [0], [0, 4], [0], [2]] Step-by-step DFS traversal: Start at 0 → Output: [0] Go to 2 → Output: [0, 2] Go to 4 → Output: [0, 2, 4] (no more neighbors → backtrack) Back to 2 → done → back to 0 → visit 3 → Output: [0, 2, 4, 3] Back to 0 → visit 1 → Output: [0, 2, 4, 3, 1] ✅ Final output: [0, 2, 4, 3, 1] 💡 Key Points: DFS is naturally implemented using recursion or a stack. For this problem, neighbor order matters — follow adjacency list sequence. Time Complexity → O(V + E) (each vertex and edge visited once). Space Complexity → O(V) for visited array and recursion stack. 🏷 Hashtags: #gfg160 #geekstreak2025 #GeeksForGeeks #DSA #ProblemSolving #Graphs #DFS #Python #DailyChallenge GeeksforGeeks

6

Databricks is heading to AI Council in San Francisco, May 12-14, with three sessions covering some of the most pressing questions in data and AI infrastructure right now. - Nikita Shamgunov keynoting on what data infrastructure actually needs to look like in the AI era - Stas Kelvich on why AI agents need a new kind of OLTP and how Lakebase is built for it - Robert Martin-Short leading a workshop on systematic LLM prompt optimization with DSPy Find us in the Expo Hall if you're there!

173

7 Comments

Last week, dbt Labs launched the public beta of dbt Fusion, a Rust based engine rewrite that brings a series of neat new features to our favorite data transformation tool: 🛠️ Real-time SQL validation and autocompletion 👀 Live previews of CTEs and compiled code 🔍 Rich model lineage at the colomn and table levels ⚡️ 30x faster project parse time (thanks Rust 🦀) 💻 Deep VS Code / Cursor integration At Opsima, dbt is a key component of our data stack (along with Prefect and Snowflake). dbt powers all data transformations across our cost and usage models and commitment automation engine. And being Cursor power users, the idea of a faster, smarter model development workflow within our IDE is quite exciting! That said, the reaction hasn’t been all positive. Reddit’s on fire with concerns about what Fusion means for the future of dbt Core and open source. https://lnkd.in/e9WSFARf dbt Labs insists Core stays under the Apache 2.0 license and fully supported, but Fusion is only “source-available”, not entirely open source, and licensed to limit hosting it as a service which raises some concerns. https://lnkd.in/ekVzgxcW For teams like ours though, we’re confident Fusion will help us ship faster and smarter while keeping core flows portable. What is your own take on the launch? #dbt #dbtFusion #FinOps #DataEngineering #OpenSource

13

#AI coding #agents are changing how I evaluate vendors. A strong argument for products is if they can be configured purely through code and #API / #MCP. Not because they have their own AI features - but because they let your AI agent do the work. Lightdash is a great example. It is not the most mature BI tool, but probably one of the most AI-ready - because they made all data and configuration accessible via API and gave AI agents the context to act on it, so Claude Code can end-to-end create dashboards by modelling data in the data warehouse and then build the visualizations. If your product is not API-first, AI cannot bring in outside context or act autonomously within it. That is becoming a real competitive disadvantage. https://lnkd.in/e6Gp3MGj

53

14 Comments

Really exciting to see this collaboration come to life. In this video, Drew Henry (Arm) and Dr. Jewel L. (Tensor) walk through what it takes to power an AI-native vehicle, from real-time perception to safety-critical systems. This follows Arm and Tensor Auto collaboration on the compute foundation of Tensor’s personal Robocar, supporting Level 4 autonomy as it scales toward commercialization in 2026. Physical AI is moving into the real world and it’s incredible to see the Arm compute platform enabling it. https://okt.to/vCMmQg

162

1 Comment

"NVIDIA founder and CEO Jensen Huang and U.S. Secretary of Energy Chris Wright announce a landmark collaboration with Oracle to build the U.S. Department of Energy (DOE)’s largest AI supercomputer to boost scientific discovery." https://lnkd.in/eBBcnmrz

216

Continuing my AI journey with UC Berkeley – Module 2: Fundamentals of Statistics and Distribution Functions. Here were some key takeaways: - Differentiated between normal and uniform distributions - Evaluated measures of uniform distribution using SciPy - Created and interpreted visual plots: PDFs, histograms, scatter plots, pair plots, correlation matrices - Discussed real-world applications of the central limit theorem - Computed mean and variance of univariate Gaussian distributions - Applied the law of large numbers to population data - Explained covariance and correlation in multivariate distributions - Plotted multivariate datasets for deeper insights - Computed measures of central tendency using pandas - Interpreted a sample covariance matrix with Python - Identified correlations from scatter plots and conditional probabilities This module reinforced how statistical foundations are the backbone of modern data science and AI. I look forward to applying these concepts in real-world projects and continuing to build expertise as my journey unfolds. #UCberkeley #Statistics #DataScience #AI #LearningJourney #Python #SciPy #Pandas

9

This is what Lucas Thelosen recently told me on my podcast—and this realization let him to treat data as a product, and adopt a producty approach to building data-driven solutions. “Once I realized that our second product is data and I started to treat it like a product, that was a pivot moment for our organization.” This was a very memorable story to me as it’s rare that I hear somebody having such a distinct moment where “the lights went on.” In his words: “It really was a wake-up moment. I was working at a tech company as the head of data, and we had invited this CEO of a very successful company to come in and talk to us. He was going through his journey — it was during the financial crisis when he took over as CEO — and he said, ‘What I realized is that we have two products: our main product, but then our second product is actually data.’ Once he realized that their second product was data and started to treat it like a product, that was a pivot moment for the organization. All of a sudden, things started to really change. He was able to sell the company for over a billion dollars to Oracle a couple of years later. I went out of this — I actually wrote him a thank-you note ten years later because it was such a crucial moment in my career. I said to myself, ‘You know what, data is the second product of my company.’ Our first product was the technology we were selling, but the second product was data, so I should treat it like that. I sat down and I wrote a plan.This is how I’m going to treat data.This is our roadmap. I was not a product manager, so I looked up best practices for product management and I presented them to our leadership team. I said, I know every day you’re asking me and my team to do certain types of analysis, but here’s how I see data as a product. This is how we’re going to tackle each department. We’re going to have six-week cycles and we’re going to move through them. We’re going to have release notes and all these things. And they said, ‘Yeah, go for it.’” – Lucas Thelosen 👇 Excerpt from episode 181 of Experiencing Data, one of the top 2% of all podcasts in the world. 🔈Listen here: https://bit.ly/47ufC6u

6

1 Comment

The AI Research Team at Cisco Central AI Org keeps innovating how LLMs can be applied to help customers in managing large complex operational problems. This blog that the team just published is a great example of addressing the 'Large Context Window' bottleneck, delivering a scalable framework that slashes LLM latency and operational costs without compromising precision. https://lnkd.in/giFG45Cx

61

Databricks and NVIDIA introduce Genesis Workbench, an open source blueprint for running modern life sciences and drug discovery AI workflows on Databricks. Learn how open source biology models and NVIDIA’s accelerated BioNeMo stack can be used to run large-scale inference, fine-tune models on governed enterprise data, and move drug discovery experiments from the wet lab to in silico—faster, cheaper, and at scale. Genesis Workbench is open source and available today!!! This is a big deal for #lifesciences and #drugdiscovery. Scientists can discover drugs faster. They can fail fast, iterate faster, and move promising drug candidates forward with confidence. Faster. Cheaper. At scale. Srijit Chandrashekhar Nair Mayank Tahilramani Adam Crown Mark Lee Oliver Chiu Prasad Kona Nima Badiey Joel Caruso

34

6 Comments

Introducing "Building with Llama 4." This short course is created with Meta, and taught by Amit Sangani, Director of Partner Engineering for Meta’s AI team. Meta’s new Llama 4 has added three new models and introduced a Mixture-of-Experts (MoE) architecture to its family of open-weight models, making them more efficient to serve. In this course, you’ll work with two of the three new models introduced in Llama 4. First is Maverick, a 400B parameter model, with 128 experts and 17B active parameters. Second is Scout, a 109B parameter model with 16 experts and 17B active parameters. Maverick and Scout support long context windows of up to a million tokens and 10M tokens, respectively. The latter is enough to support directly inputting even fairly large GitHub repos for analysis! In hands-on lessons, you’ll build apps using Llama 4’s new multimodal capabilities including reasoning across multiple images and image grounding, in which you can identify elements in images. You’ll also use the official Llama API, work with Llama 4’s long-context abilities, and learn about Llama’s newest open-source tools: its prompt optimization tool that automatically improves system prompts and synthetic data kit that generates high-quality datasets for fine-tuning. If you need an open model, Llama is a great option, and the Llama 4 family is an important part of any GenAI developer's toolkit. Through this course, you’ll learn to call Llama 4 via API, use its optimization tools, and build features that span text, images, and large context. Please sign up here: https://lnkd.in/gXKeipht

2,830

111 Comments

Multimodal AI pipelines across images, video, audio, and 3D data are increasing the importance of scalable annotation and storage infrastructure. As datasets grow in size and complexity, the ability to efficiently process, structure, annotate, store, and manage data throughout the AI lifecycle has become foundational to improving model quality and operational scale. https://lnkd.in/eecmmg9c #dataprocessing #bigdata #computervision #cvat #cloudstorage #datapipeline

8

YellowDog.ai just set a 10x benchmark uplift in scale computing, delivering 40,000 tasks per second (TPS) and managing 100,000 compute nodes in the cloud. What's even more interesting, that's 2x IBM Symphony and opens an intriguing pathway for these until-know closed/captive systems.

11

When evaluating LLMs, most teams focus on model capabilities. Our new data reveals why they should be testing function calling performance: Key Findings: → 11% accuracy gap in basic API calls between platforms → 3x more failures in chained function workflows → Output consistency varies by up to 20% The infrastructure handling your function calls matters as much as your model weights. Swipe through to see why function calling matters 👇  🔗 Read more in our blog: https://lnkd.in/gt_MFkCP 

16

Our portfolio company, TileDB , announced a partnership with Databricks to harness multimodal data to power AI in healthcare and life sciences. https://lnkd.in/g9Z56Bwk

6

What are the most underrated Databricks capabilities? Watch my conversation with Shiv Trisal (Global Industrials GTM @ Databricks) on the topic below. For the full, practical, no-fluff discussion on real-world data + AI: https://lnkd.in/gMPsFHbY

89

Last week, we added OpenAI GPT‑5.2, Cohere Rerank 4.0, and new open-source models with low-cost customization to Microsoft Foundry—giving you more choice than ever. And by now you know that while models make a difference, you need more than models to deliver business value. You need a trusted enterprise AI system that brings all the critical pieces together so you can: ✓ Connect your enterprise data so answers are grounded and compliant ✓ Automate complex workflows to accelerate outcomes ✓ Operate with confidence through built‑in security and governance ✓ Optimize cost and performance with intelligent model routing and observability ✓ Scale globally with resilient infrastructure to reach every customer, everywhere We built Foundry on Microsoft Azure to be that system so you can run AI as an integrated part of your business, not a collection of parts, and deliver results at scale. Learn more: OpenAI GPT-5.2: https://lnkd.in/gGMkVQQN Cohere Rerank: https://lnkd.in/gS-H78CR Fine-tune new OSS models: https://lnkd.in/ghusCtJg Microsoft Foundry: https://lnkd.in/gADEnhZ7

75

3 Comments