Atlan

Four internships in two years. By most career advice that's at least two too many. It's been a month since I got converted to Software Engineer I at Atlan — after stints at KoinX , Tunable Labs , and ezAIx Inc. Took me a while to figure out what I actually wanted to say about it. The thing nobody tells you: sampling broadly is the only way to actually find out what kind of engineer you want to be. Each of those four places rewrote what I thought I wanted next. Atlan is the first one that stuck. Two specific thanks: Surendran Balachandran , who showed me what serious engineering work actually looks like up close, and Mustafa Hasan Khan who's been in my corner since before I started college. If you're a third-year debating between one "safe" internship and trying a few different things — try a few different things.

Last week was pretty incredible & inspiring for a couple of reasons. 1) Atlan was at the Gartner Summit and the message was loud and clear: "Context is the brain for AI, the next critical infrastructure." 2) I also had the fortune of hanging out with the exceptional team that is building the Context Layer for AI Such an incredibly important problem statement to work on, such an interesting time to build and such an amazing bunch of humans to build with 💪

I’m the only person at Atlan allowed to update one context repo. If something goes wrong with it, every AI agent in our company is affected. That repo is our core brand voice skill. It feeds the agent on our website, the one running outbound campaigns, the one building landing pages. Our social media lead built her own channel-specific voice skill on top of it. If I update the brand voice, her agents inherit the change. Skills teach AI how to do things. They're atomic, reusable, and compounding. Our web team built a skill to analyze performance data and traffic. One of our design leads built a "creative director" skill that critiques ad and social creative. A website skill depends on a brand voice skill, a content skill, and an SEO skill. Each one builds on the others. We quickly realized we needed governance around how skills get created and updated. Who has permission to update a skill that feeds dozens of agents? What happens when one change cascades downstream? We even have someone on the ops team whose job includes keeping tabs on every skill in the org. He built a skill that spots duplicates before they create conflicting instructions. But skills alone aren't enough. They need knowledge to draw from: brand voice, ICPs, competitive intel, playbooks, customer stories. Skills are how to do things. Knowledge is what to know and why now. Together they form our context layer. Shared, version-controlled, compounding. Populated by humans and agents together. The reason we built all of this is to teach AI how our marketing team actually works. It's not perfect yet. But every week we get a little closer to becoming a truly autonomous marketing organization. What skills are you building? Curious what others are seeing.

The hardest part of scaling AI agents isn't building them. It's making sure they all agree on what the data means. At one agent, it's manageable. At ten, you start noticing inconsistencies. At fifty, you have a governance problem that impacts your ability to scale. Databricks and Atlan are partnering on a session May 20 to walk through how to build the context layer underneath your agents. Databricks Genie for querying and reasoning. Atlan for the semantics, quality signals, and policies that keep it all trustworthy. Worth joining if you're past the pilot phase and into "how do we actually scale this." https://lnkd.in/g8GJTJWd

"Where does this number on my quarterly report come from?" It's the question every finance analyst asks. It's also the question every AI agent — including SAP Joule — needs to answer to be trusted with enterprise decisions. In most SAP shops, the answer is locked inside InfoObjects, ADSOs, CompositeProviders, BEx Queries, and the Transformations between them — and most data catalogs leave that world a black box. Today, we're unlocking it. The SAP BW/4HANA connector is now in Private Preview on Atlan. Combined with our SAP ECC, SAP S/4HANA, and SAP HANA connectors, customers can trace data from source SAP application tables, through Transformations, ADSOs, and CompositeProviders, into the BEx queries that consume them — with field-level lineage through Transformation rules. But the bigger story isn't lineage. It's context. Every InfoObject, ADSO, and Transformation Atlan extracts becomes part of a rich context repository — the semantic layer AI agents need to reason over enterprise data without hallucinating. Joule. Copilot. Glean. And the many agents now reaching production. The timing matters. SAP data is no longer trapped inside SAP. Through zero-copy sharing, SAP Business Data Cloud now federates directly into Databricks, Snowflake, Google Cloud BigQuery, and Microsoft Fabric — no ETL, no duplication. But zero-copy moves the bytes. It doesn't move the context. Without lineage back to the source SAP transaction, the meaning of a custom InfoObject, the owner of an ADSO — a federated BEx result in Snowflake or Fabric is just a number, and any AI agent reasoning over it will hallucinate. That's where Atlan fits. A single context layer across SAP and non-SAP, traveling with the data wherever zero-copy takes it. More SAP coverage is queued behind BW/4HANA: → SAP BW on HANA — classic NetWeaver landscapes still in production → SAP Datasphere — the next-gen SAP data fabric → SAP Analytics Cloud — the consumption layer Once they all ship, every AI agent — Joule, Copilot, and the ones being built inside Databricks, Snowflake, BigQuery, and Fabric — will have a navigable view from source SAP applications, through every modeling layer, into the SAC stories where users consume it. The context repository for enterprise AI, end to end. Read the release notes in the comments. #SAP #BW4HANA #SAPJoule #ZeroCopy #EnterpriseAI #SAPsapphire

One thing I love about working at Atlan — we don't just talk about things, we show them. The "context layer" has become one of the biggest buzzwords in data and AI this year. Everyone claims to have one. On April 29, we're going to build one live. From scratch. No slides. Just the product. If you've been hearing "context layer" everywhere and still can't picture what one actually looks like in production — this is the event. April 29 | 11 AM ET | Virtual https://lnkd.in/ewMcWcZK

A few weeks ago, I got the opportunity to lead Atlan 's first Context Agents Accelerator, where for 2 weeks, I spent every day hand-in-hand with some amazing data teams across the world. What we built together is extraordinary, the kind of foundational work that changes what an entire industry can do next. For as long as we can remember, building context for data has been a human-only job: write descriptions, document tables, define metrics, review, repeat forever. That era is ending, and these last two weeks were the proof. The premise was simple: a 2-week sprint where the cohort got early access to our Context Agents to go from sparse metadata to AI-ready data, live for their end users, both humans and AI. Humans set direction. Agents do the heavy lifting. In 14 days, the cohort produced 1.03 million metadata enrichments, 15,000+ READMEs, and 25,000+ SQL Intelligence enrichments. 87% said the quality was on par or better than what humans would have written. Context that never existed before is now live, for humans and AI agents alike. Someone in the cohort joked that now human written descriptions feel “vanilla” and asked if they can overwrite it 😂 Meeting these teams everyday the last 2 weeks, sharing laughs, asking great questions and exchanging advice has been one of the best parts of the year for me so far. So full of gratitude and excitement! We've been working toward this future for a long time. This isn't just a glimpse of what's possible, it's proof that it's actually happening. These teams didn't just move faster. They proved that building a context layer at scale, in a matter of weeks, is finally a reality. On April 29 at Atlan Activate, we're unveiling everything Context Layer: Context Engineering Studio, nine new Context Agents, and the architecture for teams ready to build at scale. I'd love to have you there: https://lnkd.in/eJCdzzAW Massive shoutout and thank you to our Context Pioneers for trusting and building with us 💙 Kenneth Jebjerg Tish Tenorio, ARM-E, CRMP Sayali Avalakki Swatilekha Saha Bernie Daley Keith Guyett Amulya Sagi Lexie McGillis Izabela Wilczyńska Moses Ikeakhe Cody Brees 🌤️ Robert Loesch Adrianna Clark Terry Neyens Vineet Shukla Catherine Mannell Sonal Junnarkar Simon King Ethan R. Bjørn Ouni Prakash Kewalramani

We’ve reached a tipping point. For the last decade, building context for your data has been a human-only job: Write descriptions. Document tables. Define metrics. Review. Repeat forever. But now? AI can create 1 million context descriptions in three weeks. That's 4x more than humans created in an entire year. We kicked off our first Context Agents Accelerator program in March. Customers got early access to our Context Agents with the goal of going from sparse metadata to AI-ready data, live for their end users. Humans set direction. Agents do the heavy lifting. We expected a handful of teams to join. We got 50. The results still blow my mind: 1.03 million metadata enrichments, 15,000+ READMEs, 25,000+ SQL Intelligence enrichments. At the rate humans had been generating metadata, this would have taken 2.8 years. It took 14 days - a 58,000X increase in output per team. And 87% of the cohohort said the quality is on-par or better than context humans would have generated. Context that never existed before is now live, for humans and AI agents alike. We’ve been working toward this for a long time. This isn’t just a glimpse of what's possible - it’s proof that it’s actually happening. These teams proved that building a context layer at scale, in a matter of weeks, is finally a reality. On April 29, at Atlan Activate, we're unveiling everything Context Layer: Context Engineering Studio, nine new Context Agents, and the architecture for teams ready to build at scale. I'd love for you to be there for this one. https://lnkd.in/dGKP-jhu Thank you to our Context Pioneers - the first cohort of the Context Agents Accelerator program - for trusting us, building with us, and sharing your stories. Kenneth Jebjerg , Tish Tenorio, ARM-E, CRMP , Moses Ikeakhe, Sayali Avalakki , Vineet Shukla , Juan Francisco Gonzalez Sanchez , Terry Neyens , Konstantinos Tsitsirigkos , Lexie McGillis , Swatilekha Saha , Adrianna Clark, Ryan Uyehara, Robert Loesch , Bjørn Ouni , Eduardo Almada , Sergio Aceves Azuara , Amulya Sagi , Simon King , Catherine Mannell , Bernie Daley , Ethan R. , Izabela Wilczyńska , Keith Guyett , Cody Brees 🌤️ , Sonal Junnarkar

Another Atlan mic-drop moment. 1 million AI-generated descriptions. 110,000+ hours of manual effort saved. All in just 3 weeks. 🤯 Oh, and this isn't "AI makes a good first draft!" We're at the point where customers are asking if our Context Agents can overwrite their human-created descriptions 😳