Naval Katoch

The illiterate of the 21st century will not be those who cannot read and write, but those who cannot learn, unlearn, and relearn. Investing in human capability is the key to thriving in an AI-powered future.

1

Quantization in Large Language Models (LLMs) As LLMs scale, intelligence is no longer the bottleneck — memory, latency, and cost are. Quantization turns out to be one of the most powerful (and misunderstood) techniques enabling real-world deployment of large models. In this blog, I break down: • Why LLMs are surprisingly quantization-friendly • INT8 vs INT4 vs NF4 — what actually works in practice • How modern architectures absorb low-bit noise • Where quantization fails and how to design around it Quantization is no longer an optimization step — it’s foundational infrastructure for production LLM systems. Link: https://lnkd.in/gqGje75G #LLMs #AIEngineering #Quantization #MachineLearning #GenAI #MLOps

9

2 Comments

The initiative introduces a vendor-neutral semantic model specification designed to create consistency in how business logic is defined and shared across AI and business intelligence applications. #dbtlabs #snowflake #blackrock #salesforce https://lnkd.in/dKs8h8Tg

14

I’ve seen this many times: if the answer to “Why are we moving to Databricks?” is “to modernize,” then the real transformation hasn’t started. Modernization isn’t a goal. It’s a by‑product of unlocking business value. The conversation should begin with: What revenue, cost, risk, or productivity benefits are we targeting? Without that, adopting a new platform is just accumulating technology. Once value is explicit, sponsors understand that while new use cases deliver visible benefits, there is also essential “invisible” work, e.g.: migrating logic, removing duplication, and retiring legacy patterns....to avoid mismatched numbers and parallel stacks. From experience, a practical approach is: - Bring all data - including transformed reporting tables - into one place to avoid the “system X vs system Y” issue. - Stop new work on legacy systems and govern exceptions. - Migrate and build new capabilities in parallel, with clear focus if the team is small. - Keep legacy‑derived tables in the new platform for backwards compatibility. - Educate continuously: analytics is an iceberg—value is visible, but most effort sits below the surface. If business value leads, modernization follows. If it doesn’t, you’re just adding another platform.

13

2 Comments

Another fantastic blog from Ganesh Bharathan on query optimization for Snowflake! Read his analysis with example code below! In the blog he provides a deep dive into the following topics: 🏆 Data modeling and query compilation 🏆 Concurrency and warehouse sizing 🏆 Join explosion 🏆 Large table scans 🏆 Data retention and archiving 🏆 Disk spillage 🏆.Complex queries https://lnkd.in/ergJqeQf

23

1 Comment

I am going to delete this post. Biryani is still cooking. I will publish a detailed one. usecases: 1. Create Partitioned tables from Spark. \Read and Write from Snowflake and Spark 2. Create Partitioned tables from Snowflake. \Read and Write from Snowflake and Spark Ashish Mishra. Thank you for checking. ✔️ From snowflake: \create partition table in polaris catalog linked database ✔️ From Spark: \I can test the Hidden partition feature for data inserted from Spark. 😡 \Reading snowflake inserted data failed    spark.sql("select * from polaris.market.CUSTOMER_ST ").show() Caused by: java.lang.UnsupportedOperationException: Cannot support vectorized reads for column [ST] optional binary ST (STRING) = 4 with encoding DELTA_LENGTH_BYTE_ARRAY. Disable vectorized reads to read this table/file Biryani: The mutton at the bottom of the biryani dish was not cooked properly. Snowflake https://lnkd.in/gDteWqb2 CREATE ICEBERG TABLE ambari_prod_LINKED_DB."market".customer_st (  first_name STRING,  last_name STRING,  amount INT,  st STRING ) PARTITION BY (st);    INSERT INTO ambari_prod_LINKED_DB."market".customer_st VALUES ('Satya22', 'kondapalli',100, 'IL'); INSERT INTO ambari_prod_LINKED_DB."market".customer_st VALUES ('Satya21', 'kondapalli',100, 'IL'); INSERT INTO ambari_prod_LINKED_DB."market".customer_st VALUES ('Satya31', 'kondapalli',100, 'WI'); All data files are created in the 'data' directory. What is the purpose of a partition?"

10

6 Comments

Claude, ChatGPT, Snowflake Intelligence, Cortex Analyst, Databricks Genie, Zenlytic, TextQL. Think about any AI agent you connect to your data. They all get questions from business users like: “What was the average number of daily active users in the last quarter?” These AI agent connects to Euno’s MCP server to retrieve metadata. That’s when Euno gives these agents the missing context they need to understand your data. ** Just a reminder that Euno is a metadata platform so we only access your metadata ** Euno runs a semantic search and looks for dashboards that use DAU (which is semantically equivalent to Daily Active Users). It filters for certified dashboards using the trust signals you’ve already set in Euno. In this example Euno finds two candidates: 1. DAU used in the User Engagement dashboards 2. DAU_NEW used in the Product Statistics dashboard Both dashboards are certified. But Euno also shows how often each has been used in the past 30 days, along with the owner or creator. With that information, the business user can confidently choose the right one and the AI agent can deliver the correct answer. The same applies to tools like Cursor. Let’s say you’re an engineer and ask: “Can you find me the Tableau dashboard for Customer Success?” Cursor will reach out to Euno’s AI assistant to get that BI context because Cursor is focused on dbt development and doesn’t have access to that layer. You can even go further and ask which dbt models power that dashboard and how those models are calculated. Euno feeds Cursor with that real-time context as you build. It’s not just for business users. It’s for developers too.

25

5 Comments