Rishi Bhatnagar

The 'universal semantic model' getting lots of attention as the path forward for AI(LLM) to leverage your data stack in a more governed/controlled way, and get your common business logic out of the BI semantic layer. Becoming a more critical topic of conversation as data teams/CIOs/CTOs are looking to position their stack for the future, and seeing many players, old and new (dbt, cube, honeydew, atscale, apporchid, ...). Gordon Wong Joe Reis Benjamin Rogojan -- what are your thoughts and what are you seeing?

13

2 Comments

Are you struggling to get a data governance program off the ground? Are you having difficulty getting your business stakeholders aligned around specific data definitions and quality standards? Or, perhaps you struggle to get your customers fully engaged in data stewardship activities? If yes, I have some advice. This advice comes from the school of hard knocks - having been in several situations where I wrestled with all the issues highlighted above. Here is my advice: for early stage governance efforts, wherever possible, look to use 𝐭𝐡𝐢𝐫𝐝 𝐩𝐚𝐫𝐭𝐲 𝐫𝐞𝐟𝐞𝐫𝐞𝐧𝐜𝐞 𝐝𝐚𝐭𝐚 as a source of 'truth'. Why? ✅ One of the greatest values of third party data, especially if you buy from a reputable company, is that it comes 'pre-governed'. Your vendor does the hard work of defining and enforcing governance standards for that data. ✅ Third party data is generally easy to 'defend', especially when it comes from a reputable vendor. So, when your internal customers look at a report and ask 'where did this come from?', it's easy to explain what they are seeing when it comes from a reputable vendor in the space. ✅ Your 'time to market' of any data management solution that relies on a source of truth, like MDM, will be significantly reduced. Instead of spending days in governance committee meetings debating definitions and quality standards (assuming you can even get people to attend), you can be focused on solving specific business problems. ✅ Your need to steward any data provided by a third party should be negligible - assuming the governance standards of the vendor well-align to your existing (or expected) governance policies. These are some of the many benefits of using third party data, but with the benefits, come costs. Often, these costs can be significant. Worse yet, its impossible for a given data buyer to know if what they are paying for third party data aligns to industry standards, or if similar quality products exist on the market at a similar price. This lack of transparency on the price and quality of data providers is exactly the problem that my friends at Blue Street Data exist to solve. So, if you see the value in third party data as an effective tool to accelerate your data governance efforts - and need help in the process of engaging a potential vendor - then I recommend you check out the buyers guide linked below. It provides you detailed guidance on every step of the process to acquire third party data, and will help ensure that you are getting the right product, from the right vendor, and the right price. Check it out! #thirdpartydata #referencedata #datagovernance https://lnkd.in/eHCAsU9s

28

1 Comment

MDS Fest 3.0 brought together over 60 speakers last week for its third annual virtual data conference including talks about data leadership, data storytelling, and the impact of AI. Here’s one quote and two key insights from three of my favorite talks from the week! Managing icebergs (not that iceberg) — Jake Hannan • “Every team has their own set of icebergs but you can build processes and shared knowledge around your work with your stakeholders to ensure you don’t end up sinking” • Data teams have to manage an iceberg consisting of both visible priorities above the water, such as topline deliverables for the organization, as well as complexities hidden below the water, such as maintaining databases and managing the accumulation of tech debt • Sigma’s data platform team manages their iceberg by structuring the team into functional pods (GTM, Product & Engineering) to closely align analytics engineers with relevant stakeholders and by using a scoring matrix (impact, confidence, and effort) to prioritize requests for the data team in order to maximize downstream impact Behind the build: ClickUp’s data platform — Michael Revelo • “Don’t forget about the human side of the data platform… a lot of the actual SQL modeling is going to be taken over by AI so a lot of the advantages are going to be understanding the business and activating the data” • Simple things like standardized syntax are crucial when scaling the business because they improve understanding of the data warehouse, especially for non-data team members, by creating an “implied data dictionary” (e.g. all of ClickUp’s intermediate dbt models are for filtering out bad data) • Keeping a pulse on the data startup ecosystem enables you to identify, partner and grow with early companies (e.g. recent Y Combinator cohorts) in a symbiotic relationship that can also benefit your build vs. buy balance by leveraging those relationships in the future How do you measure the ROI of your data team? (Spoiler alert: You can’t) — Julia Slisz • “You can ask a lot of questions but, at its core, whether people like what you’re doing and want to work with you again in the future is the ultimate way of knowing if you are adding value” • The real measure of ROI is NPS (net promoter score) a.k.a. your customer’s willingness to recommend you to others • The best data teams make speed a priority (i.e. tight feedback loops and timelines really matter), win friends and influence people (i.e. having people believe your empathy for their business situation), and focus on the “new, the novel, and the gnarly” (i.e. instead of focusing on quick questions, they focus on what they can uniquely do to answer the most strategic questions facing the business)

7

Excited that Kurve now integrates with Snowflake Semantic Views—making it faster to define relationships between entities and embed critical business context!

13

wrapping up a 6mo project with a large client in the enterprise manufacturing space this week. we migrated materialized Snowflake views to a new dbt Labs project: - 322 staging files - 71 intermediate tables - 35 mart tables - 7 incremental tables I used the opportunity to lean into new AI code dev tools: - Anthropic's claude code within Cursor - nao Labs (YC X25) the client was ~18k person org. im proud of the rate at which we got things done despite the friction we faced in (2) areas. (1) our end product was affected by compromises on quality to appease the client who wanted things a certain way / didn't prioritize certain dev standards, in lieu of faster code deployments (lots of small things e.g. no lint checks in CI jobs, bypassed code approvals). a consensus on compromised quality is toxic to a project. It’s a silent agreement that discourages individual contributors from pushing a task closer to perfection, once they know their colleagues won’t be. It kills motivation. The collective end result is haphazard work and tech debt. (2) client-side analyst/engineers became defensive over recommendations and hard to work with. in these situations I try to stay firm on my position while being mindful / proactive addressing personal feelings at stake. work is emotional and people are sensitive. I think its better to have an uncomfy chat about feelings hurt than allow hidden resentments to fester eg 'im sorry you feel belittled by this choice....im pushing for X to achieve Y but can see how its causing Z and thats not cool'. vibes impact quality always learning. headed to asia for the next 2mo. stay tuned massive thanks to my consulting colleagues. (i wish you could add spotify songs to linkedin posts 😅)

53

6 Comments

👀 Entity resolution is a hot topic across most enterprises. The Databricks + The Trade Desk partnership/integration will make this easier for a lot use-cases. 💪

11

The cost associated with not focusing on strategy for your ontology is bigger than you think. Let's augment the FDE model with strategic thinking going forward. We'll get better, non-linear outcomes and reduce the churn around development in Foundry and AIP. CodeStrap's "Code and Connor" Episode 13 releases on Monday, September 29nd, 2025 with our friends at Foxtrot Services: co-CEOs Christine Williams and Nicole Sanders and CTO Martin Seebach. Foxtrot Services are field-proven Palantir Foundry & AIP engineers with 40+ secure, large-scale deployments under their belt and growing. Our thirteenth episode focuses on: → The buyers of AI are the business → Governance, Security and Observability around AI → Ontology as a product → The new security threats posed from AI → AI has been the forcing function for companies to get their data in order → The generational and historical setting for AI adoption → IT vs. FDE → Unpacking the FDE role → AI success stories and building trust → The explosion of data and AI coding may lead to the Codepocalypse → The cost associated with foregoing strategy is bigger than people think Be sure to follow us to get the latest information, and schedule a meeting with us through www.codestrap.com!

21

The organizations that will execute on AI will be the organizations that have the capability to scale quickly as any AI innovation happens. Open AI just announced GPT 5.2, and not only is Databricks the only platform that already supports it, but we used our benchmark to measure it against enterprise tasks.

23

1 Comment

Databricks and Perplexity co-founder Andy Konwinski is forming the Laude Institute, a new AI research institute, and backing it with $100 million of his own money. The Laude Institute will function less as a traditional research lab and more like a fund making investments structured as grants. It will focus on "Slingshots" (early-stage research) and "Moonshots" (long-horizon labs tackling species-level challenges). The institute's board includes AI luminaries like UC Berkeley's Dave Patterson, Google's Jeff Dean, and Meta's Joelle Pineau. The institute announced its first flagship grant of $3 million annually for five years to anchor the new AI Systems Lab at UC Berkeley. The new fund aims to catalyze work that doesn't just push the field of AI forward, but guides it toward more beneficial outcomes. Congratulations to Andy Konwinski on this incredible commitment to AI research. TechCrunch: https://lnkd.in/dy7x9zAT #AI #Research #AIforGood #Philanthropy #VenturePhilanthropy #Databricks #Perplexity #UCBerkeley

2

Almost every CDFI leader I speak to -- and I speak to a lot -- wants to do more with their data. What holds them back is not a lack of motivation or ambition ... it's a lack of resources. For these community-rooted investors, AI can make a huge difference in reducing costs, saving time, and deepening impact. But it all starts with data. And Pacific Community Ventures (PCV)'s Radiant Data Hub is here for it. We'll wrangle your data and get you AI-ready. When I met Tamra Thetford at CNote, our connection was instant. This article almost wrote itself the more we chatted! A huge thank you to the funders out there who get it, and support this work every day - Erica Trejo, Anina Tweed, Vilas Dhar, Nick Cain, Antoinette Marie, Victor Burrola. If you're a CDFI with a vision for better data and ethical AI, we'd love to hear from you ... what's holding you back?

40

Your AI Tools analytics are only as good as to the accuracy of your data. Certify your data sets with Saras Analytics

7

Good news for ALL customers trying to create semantic models & the context layer for their AI tools (like Snowflake Intelligence/Databricks Genie/ and others) to work. We are Snowflake native on their marketplace & have started customer deployments. 🚀 AI needs more than metadata. It needs: - Semantic relationships between entities, - Business definitions in natural language,  - Measure definitions and calculations logic, & - Domain specific context and terminology - the enterprise brain - Agentic systems to understand data structures Creating all this takes 3-6 months/ sometimes longer - stalling AI adoption. At ekai, we produce all the artifacts that downstream AI applications need: data catalogs, business glossaries, metrics definitions, lineage maps, validation rules. The entire context layer, generated and maintained automatically. And we can serve it. Shahid Azim Patricia Geli Beth Porter David Berlin Kate Landmann Lucky Byas Yaser Najafi Ryan Lieber Niels ter Keurs Anahita Tafvizi Eddie Blackwell, MBA, MSLC Wayne Wilson

50