Jacob• Adler
Santa Monica, California, United States
2K followers
500+ connections
2K followers
500+ connections
About
Data and AI consulting at Riverboat Consulting (riverboat.ai). I work with ambitious…
Activity
2K followers
-
Jacob• Adler shared thisExcited to share a new case study published on my marketing site from working with the Disco team this spring. We touched all parts of the data stack but focused on agentic analytics with Hex AI agents. Saw usage from the business skyrocket in the few weeks after we added basic context files. Great working with Peter Li, Matt Organisak, Deepika Singh, and everyone else! Super cool working with an AI-forward organization like them. Check it out! Link is in the comments.
-
Jacob• Adler posted thisA mistake I see companies make after they hire analytics engineers? They try to shift all the analytics engineering work over from the data analysts. It's true that you want to clear the path to doing the work that has commercial impact. But writing your own dbt code is something that can empower analysts and help get results faster, rather than putting in a ticket to analytics/data engineering. dbt shouldn't be too technical to pick up on the job — if you know SQL, you can learn dbt. IMO it's silly to retcon a tool designed for analysts as now too technical for them. Besides my belief that it's just faster for analysts to work in dbt, it's also a risk for them to not. Like I said in my post yesterday, it's crucial to understand the lifecycle of data. If you own a metric or dashboard, your stakeholders are coming to you with questions. You'll look much better if you can assess what happened and how to fix it when something goes wrong. What companies need to do in this situation is align on the right delineation between data analysts and analytics engineers. This will vary between companies based on current staffing (and skillsets) and the nature of their data. If I had to generalize one rule, it would be that analytics engineers handle the foundational data marts used between domains as well as the most technical projects. Then analysts own domain-specific mart and reporting tables. This means you still want more analysts than analytics engineers. Way more. The rule of thumb I used to have was 4-5 analysts per analytics engineer. Curious whether that still holds.
-
Jacob• Adler posted thisFor data analysts, it doesn't take much broadening of ones horizons to make a big difference in the value you bring. Earlier in my career, I felt like I'd see a lot of desire to work on data science and ML projects, in many cases where it wasn't really needed. I'd argue for the following as a more productive way to upskill: If you lean more commercial and an analytics/data engineering team works in dbt, you should learn dbt. If you know dbt and your team uses Airflow, Dagster, or another orchestration tool, you should learn the high-level basics of that tool. If your company has a Rails app, go through some basic Rails tutorials. You're not changing the scope of your role and actually shipping new features or ETL pipelines. But just some understanding of the lifecycle of data is immensely valuable. The next time something breaks overnight and your stakeholders are hounding you, you can more quickly and accurately pinpoint the problem, instead of tagging in another team and becoming the intermediary. 3-4 years ago, I went through some simple Django CRUD app tutorials and took Dennis Hume's CoRise course Data Engineering with Dagster. I felt more confident meeting with engineering and data engineering teams after that point. It's ultimately your responsibility to understand how data gets generated and makes its way into analysis and reporting. Knowing the system inside and out leads you to stand out.
-
Jacob• Adler reposted thisJacob• Adler reposted thisLightdash is hiring a Head of Engineering! This is the hardest role we've tried to fill. If you pass these three quick tests, I'd love to chat (I'm yet to meet somebody across all 3): 1. You're an exceptional people leader. You know what a world class engineering team looks like. 2. You have technical product chops and you've built delightful products. 3. You know engineering is dramatically changing under Agentic Coding. You've already run experiments and you have a plan to radically overhaul engineering for 2026. If these sound exciting to you, Lightdash is a fast growing Agentic Data Analytics product loved by huge companies and fast growing startups. We place value on transparency, velocity and building everything with customers. If this sounds like you or you know somebody, drop me a DM or apply in the comments
-
Jacob• Adler shared thisCan't wait to try this out — hoping this cuts out a lot of the use cases where I'd normally check the dbt Docs
-
Jacob• Adler posted thisData practitioners are used to it. When you'd get a vague question from a stakeholder, it would usually take some back-and-forth in Slack or a quick call to make sure you understand what they're asking for, why they're asking for it, and whether that's the right angle to satisfy the underlying intent. AI agents aren't necessarily used to it. If these ad hoc questions are now going to AI chats instead, that layer before delivery is gone. AI agents want to answer the user's question, and without more info its answer is unlikely to be helpful. Depending on the response, the stakeholder might reply and get into a frustrating thread as the second and third agent responses still don't match what they're looking for. Or they might take that unhelpful answer and use it. This isn't necessarily a problem that's new from AI — criticisms of self-serve analytics rhyme with this. Three things your data team needs to be doing to handle this: 1. Encourage your AI agent to ask follow-up questions (in the system prompt, skill files, or anywhere else back-of-house). On a recent agentic analytics project that used Hex, one of our domain guides noted a commonly-confused question type and refused to answer until the user clarified their intent. 2. Train your stakeholders to become prompt engineers. From my experience, they've been happy to provide as much detail as the AI needs — they just want to know the best practices for what to specify. This could include what to include in their prompts like what timeframe or metrics to look at, when to push back at the AI results, when to use AI or go to the data team, or even just what context the agent has access to. 3. Stay in the loop. Whether this is monitoring chats like in Hex's Context Studio, staying in close contact with stakeholders when they're taking action based on self-serve analytics, you need to ensure that the quality of the data the business works with matches the standards you previously set.
-
Jacob• Adler shared thisThe purpose of analytics teams is to help the business make better decisions. There's a bunch of skills needed to be good at the job (particularly the technical and communication), but the best data people I've worked with had tremendous data literacy skills. Do they ask the right questions, do they spot data quality issues, and can they draw the right conclusions? There's a natural parallel with software engineering. Would you trust a key function of your business to a SaaS that was vibe coded by someone non-technical? I wouldn't have a problem with something smaller or even a one-off use case, but would worry about what would happen if my company relies on something that could be a house of cards. There could be security issues, my data gets deleted, etc. Data literacy skills don't necessarily need to come from the data team — I've worked with finance and operations folks who excel here — but like vitamins, you need to get them from somewhere. Data analytics isn't completed when you've set up dashboards, but when you've proven the system works and made a difference in helping the business make decisions.Jacob• Adler shared thisThere's a lot of speculation in the industry about role boundaries shifting and merging, thanks to AI. I am definitely hearing about and seeing this happening, albeit in less dramatic ways than folks are broadly declaring. But a pattern that seems to be emerging is people leveraging a new capability without having the supporting skills. It's easy for me to come up with examples from the data world: anyone can pull the numbers, but they don't know rules for safely sharing them. They can define their own metrics, but they don't know how to assess the quality of the underlying dataset they're defined against. They can generate any chart they can dream of, but they don't know how to interpret it. The second order consequences of this pattern are going to be rough for a lot of companies, with the Meta news today perhaps being an example of this. It feels like we're rapidly moving into the FO phase of the AI boom's FAFO push.
-
Jacob• Adler shared thisSuper exciting from the Hex team. Adding cron scheduling to Threads is a clear win, and something that will be table stakes (if not already). I love that the Tasks Cookbook highlights anomaly detection as a use case. A lot of small and midsize companies don't necessarily want to pay for observability tools when they have other data platform costs, so anomaly detection is neglected. That's dangerous — your stakeholders may find bugs before you do, or maybe you miss them entirely. This doesn't fully replace those tools but will at least replace some of the functionality for no new vendor spend.Jacob• Adler shared thisAs an important Executive Business Stakeholder™, I have a lot of questions about my business, what's going on, and which trends I should be paying attention to. A year ago, I'd go through dashboards to get started understanding things. With the Hex Agent we made it easy to just ask... but you still have to ask. Now with Agent Tasks, Hex will notify you proactively. Think of it like a "scheduled analysis". It can do simple things ("tell me what's up with this metric") or complex ("dig through all our data and build me a detailed story"). A few weeks ago, I set up my first Task and now the answer just lands in my DMs before I start my day. I don’t need to open the dashboard or type a prompt, the answer is already in front of me. Excited to see what you do with this!
-
Jacob• Adler shared thisSpent a few minutes this morning testing out the new dbt Wizard, the new coding agent from the dbt Labs team. I gave it a small task and Wizard completed it fairly easily, so no real complaints. Installation and setup were super easy with an existing dbt project. It was able to find my profiles.yml and plugging in an Anthropic API key worked the first try. I've been pretty siloed in Claude Code the last few months while the token subsidies last, so I'm a bit hesitant on how much extra I might spend instead using Wizard with BYOK (bring your own key). However, if the costs are reasonable and you're getting work done faster, it's fine. I'm not opposed to a new harness, anyway. The Claude Code team has shipped so much vibe code that Claude Code ranks fairly middling in the terminal-bench harness leaderboard, even among models running Claude. The switching costs are low enough that we can still choose when to pick each harness or migrate entirely.
Jacob• Adler commented on a post
1w
I’ve also seen with the long tool lists 1) they know a few well and then include stuff they used once 2) they’re applying for different roles and aren’t tailoring their resume to this specific one
-
Jacob• Adler liked thisWe're hiring a Lifecycle Marketer at Profound!! If you love building nurture flows, running experiments across email + in-product, and want to own lifecycle end-to-end at a fast-moving AI startup - this is the opportunity you've been waiting for 🚀 We're also hiring across the entire company 👀 . Reach out if you're interested!
-
Jacob• Adler liked thisJacob• Adler liked thisI made a skill, `tell-codex-to`. I use it to have Fable tell Codex what to do. Now that I'm fableless, well, free-Fableless (I ponied up $300 to keep using it thru usage-credits), I need to actually pay attention to how much I use. The easiest way to do that is offload menial work from Fable to a non-god-tier model like gpt5.5. Full skill in the comments, but it's basically: Delegate to Codex (default for hands-on work): - implementation from a frozen spec; refactors; mechanical migrations - bug fixes with known repro; test writing; coverage fills - CI fixes, dependency bumps, scripts/tooling - bulk codebase exploration where raw reading ≫ the answer Keep in Claude: - design, API design, architecture, naming, UX judgment - tasks where writing the spec IS the work (ambiguity = design) ...
-
Jacob• Adler liked thisJacob• Adler liked thisBig news: I've joined the Talent Acquisition team at Clay! Really excited to be joining at this impressive stage of growth, and more importantly to help build the team alongside amazing humans 🥳 Let's get it!
-
Jacob• Adler liked thisJacob• Adler liked this#1 best-selling business book in Japan for a brief shining moment. I'll enjoy it while it lasts!
-
Jacob• Adler liked thisVery cool. Nowadays I pretty much only build my pipelines with agents. Pairs really well with agents building the models and reports downstream.Jacob• Adler liked thisWe built Agent Skills so Claude Code and other AI agents can build, monitor, and deploy real-time data pipelines in Estuary. Mark Van de Wiel tested them by replicating two Postgres databases from scratch, and had data flowing in under an hour. Along the way, Claude Code caught a schema mismatch causing failures and fixed it without being told how. Agents won't replace engineers, but they will cut down the context switching so engineers can stay focused on the actual problem. Read how it went by clicking the link in the comments 👇
-
Jacob• Adler liked thisJacob• Adler liked thisI'm excited to share Joon's partnership with Ladder. I've been looking for an app that ties strength training, coaching, and nutrition into one seamless experience for a while now. Ladder nails it ✅ Weekly workout plans are built around progressive overload for real, long-term strength gains. You get in-ear coaching, video demonstrations, and integrated progress tracking whether you're at home or in the gym. The nutrition side is just as dialed. Log meals instantly by photo, voice, or barcode scan. What makes this partnership feel right is how naturally Ladder maps to two of the pillars we think about most at JOON: movement and nutrition. Both are core to why we built this all-in-one employee wellbeing platform. Proud to bring Ladder to the JOON ecosystem so more teams can access expert-led fitness and nutrition through their wellness benefits!
-
Jacob• Adler liked thisJacob• Adler liked thisReminder that data pipelines break every day, Al agents needs constant refinement, and client calls will constantly swarm any open slot on your cal, but the #WorldCup will only be in your backyard once in your lifetime. DataTropic took all of its employee to the #Spain vs #Austria match last week and it was a childhood dream come true, still can’t believe it. Also never said these words before but shoutout to LA infrastructure, getting in/out of #SoFi was a breeze thanks to the shuttles/trams/trains
Experience
-
Riverboat Consulting, LLC (riverboat.ai)
-
-
-
-
-
-
-
-
-
-
Education
-
The Wharton School
-
-
Activities and Societies: "The Daily Pennsylvanian" independent student newspaper, Alpha Epsilon Pi Fraternity. Research/teaching assistant to professors: Scott Rosner, Adi Wyner, Stewart Friedman. Intern for Penn Athletic Communications. Wrote 135 articles about sports analytics between 2014-16 for numberFire, SB Nation, and Sports Quotient.
-
-
-
-
-
-
Honors & Awards
-
Valedictorian
-
4.0 unweighted GPA, 11 Advanced Placement courses.
-
Marian Greenblatt Social Studies Award
Quince Orchard High School Social Studies Department
Given annually to a junior at Quince Orchard High School that has excelled in social studies courses during their freshman, sophomore, and junior years.
Languages
-
English
Native or bilingual proficiency
-
Spanish
Limited working proficiency
Recommendations received
1 person has recommended Jacob•
Join now to viewOther similar profiles
Explore more posts
-
Michael Brenndoerfer
Building agentic workflows? There are actually only a few key concepts to be aware of. StateGraph orchestration: A StateGraph is the agent’s blueprint. It consists of nodes that handle tasks with edges (directions) to other tasks, defining the order. Together, they provide the agent with a flow to follow. Conditional routing: At decision points, the agent inspects results and follows “if-else” statements that skip failures, handle edge cases, etc. Memory and threads: Thread memory holds recent conversation turns. Essentially, it stores the short-term memory that you are John Doe and were interested in the weather. Long-term memory stores user facts, letting the agent stay context-aware, reduce repetition, and personalize replies, such as that you have been a customer since 2018. Human-in-the-loop: Strategic hand-offs let people review low-confidence outputs, catch critical paths or errors, and feed corrections back to the model, keeping quality and alignment tight. If this still feels too abstract, I’ve included a more detailed explanation in the article below. https://lnkd.in/eAr-6YAQ
1 Comment
Explore top content on LinkedIn
Find curated posts and insights for relevant topics all in one place.
View top content