Michelle Tanco

I’ve been seeing a lot of comparisons around AI coding tools lately, and this one from Gilberto Hernandez really stood out. Same prompt. Very different outcomes. One experience required manual steps and troubleshooting to get anything usable. While the other generated a full working app, handled deployment challenges, and delivered something you could actually use. That’s the difference I keep coming back to: AI that suggests vs AI that executes. Cortex Code doesn’t just generate code—it understands the task, works through blockers, and gets to a finished result on governed data. That shift matters. Especially when teams are trying to move from experimentation to real production workflows. Curious to hear what others are seeing.

23

Excited that Cortex Code CLI is now generally available. Cortex Code is the Snowflake-native AI coding agent that turns complex data workflows into simple interactions—all in natural language. Whether you’re in Snowsight or working via the CLI, Cortex Code is deeply aware of your Snowflake environment, meaning Snowflake expertise becomes an always-on capability of the platform—for everyone. Dive in to learn more: https://lnkd.in/g-qaZJpV

13

Feature engineering is often the most critical (and complex) part of ML. This guide provides a clear introduction to Snowflake Feature Store and how to structure entities, feature views, and datasets for model training and inference.  A helpful resource for data scientists and ML engineers looking to standardize and scale feature workflows.

8

In this world of Agents.. how do you ensure that your agent is doing what it's designed to do? Agent Evaluation is one of the core capabilities required for Explainable AI. @Snowflake #Cortex #Agents Evaluation is now GA.

19

While writing a blog post explaining how #ZeroETL works and comparing it to #Kafka, I discovered they are complementary and can work together quite well to address customer's data sharing issues. If you want to leverage Zero ETL in your Kafka deployments, I help you understand how you can do it.

7

If you want to see some of the key features of Databricks' Unity Catalog in action (like ABAC, Lineage, Lakehouse Federation, etc) highly encourage you to check out my video with Paul Roome: https://lnkd.in/e9hw3rmj

41

Snowflake is releasing so many AI features lately, it's really hard to keep track of them all. For AI Agents, evaluation is very important to improve their performance. But setting up an evaluation system yourself takes a lot of work. Fortunately, the Evaluation feature for Snowflake's Cortex Agent is now in GA! If users just need to prepare evaluation data they can utilize this built-in feature to run continuous evaluations. In my blog post, I explain how this feature basically works, and also share my favorite part: how to run it and get the evaluation results using just SQL. [medium]: https://lnkd.in/gyj-Qu9n [zenn]: https://lnkd.in/gzMV2k_P

52

It’s a trope that most organizations are “data rich but insights poor.” The biggest challenge is how much *work* there is in accessing, gathering, moving, manipulating, cleansing, and then building with that data. Industry-wide partnerships like this should help lower the barrier to entry and make it easier to use this data for better decision making and quicker/more proactive action.

20

Layered Resonance Discovery Modern information discovery often requires navigating vast, heterogeneous datasets. Traditional search methods struggle with nuanced queries and identifying truly novel insights. This architecture addresses the need for a system capable of deeply understanding query intent, exploring diverse data sources, and synthesizing information with high fidelity. It aims to move beyond simple keyword matching to a more resonant understanding of underlying concepts. #AI #AgenticAI #MultiAgent #KnowledgeGraph #SemanticSearch #InformationRetrieval #DataMining #NoveltyDetection #DistributedSystems #ResonanceTheory

2

This solves a huge problem many of the agencies I work with are facing. How can we extract data, and hence insights, from our huge backlog of existing unstructured documents (e.g. PDFs). Now we can do it with both better accuracy and lower cost!

2

This is a clear example of AI infused into the data platform. Partitioning, Z order, V order... they all require knowing the shape of your data and how it will be used in advance. But you don't know. The world is being flipped on its head. Liquid Clustering adapts to use. Your data platform needs to evolve with you.

30

1 Comment

🤖 Think you know what the “R” in RAG stands for? Think again. While most people associate #RAG with Retrieval-Augmented Generation, the retrieval part often hides more complexity than we realize. In this #DataManagement blog, Terry Dorsey, Sr. Data Architect/Evangelist North America at Denodo, dives deep into what retrieval truly means — and why understanding it is critical to building effective #GenAI applications. 🔍 You’ll learn about: - The naïve RAG pattern and its limitations - Retrieval techniques for structured vs unstructured data - When to use each — and why combining them can be a game changer - Emerging new RAG patterns like GraphRAG, Agentic RAG, and Adaptive RAG Whether you’re designing #AI assistants, chatbots, or analytics copilots — this post will help you rethink how retrieval powers generation. 👉 Read “The R in RAG” now: https://okt.to/Y5kciS #LLM #DataRetrieval #DataForALL #NoDataLeftBehind #NoAIWithoutData

20

Great new launch from the Databricks AI team! We trained a document parsing system that delivers leading quality at 3-5x lower cost and outperforms frontier models like GPT-5 and Claude. https://lnkd.in/gJA8cTTi This is critical to connect AI to so many kinds of data that is locked in PDFs and other documents. We parse each document into a nice JSON format, including tables, text, and even figure descriptions, which you can then feed into AI agents or downstream apps. We also made it super easy to use from SQL or Apache Spark. Just call ai_parse_document and we automatically handle batch execution at scale. Docs here: https://lnkd.in/gCQnjF7a

2,744

69 Comments

Excited that Cortex Code CLI is now generally available. Cortex Code is the Snowflake-native AI coding agent that turns complex data workflows into simple interactions—all in natural language. Whether you’re in Snowsight or working via the CLI, Cortex Code is deeply aware of your Snowflake environment, meaning Snowflake expertise becomes an always-on capability of the platform—for everyone. Dive in to learn more: https://lnkd.in/e-3Uutic

2

It's been a while since I wrote a blog. For anyone working on agentic workflows, don't forget about observability. Helps a lot in understanding how these 🤖 are making decisions, using tools, and not to mention, save you a lot of debugging time 🔧 The article below discusses how to utilize the open-source tool, Arize Phoenix, to enhance observability in your agentic workflow. https://lnkd.in/gZBgAkhN

36

1 Comment

Simple, at scale, secured, governed, performant and cost efficient! There is a plethora of untapped unstructured data and ai_parse_document function is by far the simplest way to uncover the latent information! -> https://lnkd.in/gtct44nk

1

I presented yesterday on how we are using Databricks to automate responding to technical questionnaires (aka RFIs). It is some pretty innovative stuff driving a ton of value for our Analyst Relations and Field Engineering teams - check it out. Repo here, btw: https://lnkd.in/eaSNsYxF

11

2 Comments

🚨 Open Source or Open Confusion? A Tale of Two Strategies. It’s becoming a familiar pattern from Databricks: 👉 Launch a proprietary feature. 👉 Release a watered-down OSS version under the same name, but with different code. 👉 Market the proprietary version as if it’s the open-source one — glossing over the missing functionality. Take Unity — there's "Unity OSS" and "Unity DBX" — same name, same logo. same website for both with completely different realities. The OSS version lacks most of the enterprise features Databricks actively promotes. Yet that’s not clearly communicated. 🤷‍♂️ Contrast that with Snowflake's approach: ✅ When it open-sourced Polaris, it gave it a distinct name. ✅ It clearly communicated that Polaris is separate from the internal Horizon catalog. ✅ No ambiguity. No confusion. Customers know exactly what’s OSS and what’s not — and they can still self-host Polaris or use it as a service. And yet... we continue to see FUD-driven slides from Databricks taking cheap shots at Horizon, without acknowledging that Snowflake offers those same features via an open-source Polaris stack. Transparency matters. OSS is about trust. Let’s not dilute it with misleading narratives. #OpenSource #DataEngineering #Snowflake #Databricks #UnityCatalog #Polaris #TransparencyInTech

77

4 Comments

Excited about the recent Anthropic release of #Skills, I explored their application beyond Teradata analytical functions to educating #LLM on constructing advanced pipelines from customer input (table name). I generated a set of #Skills enabling #LLM to craft these pipelines, showcasing a process from data prep for training to evaluating end results, ensuring pipeline efficacy. This demo illustrates the potent capabilities of #LLMs with #MCP tools for intricate analytical tasks. Currently, I've developed 52 #Skills for testing, a continuing endeavor that has yielded promising outcomes. Keep in mind this is work in progress. Access the experimental skills here: https://lnkd.in/gkzGfWgj

59