Dov Schnaider

I wrote an article with David Yuan and Diego Z. from Tidemark, where we dove into how AI is shaking up the data game. For years, the big players—those with massive datasets—had an edge because their data acted like “gravity,” pulling customers in and keeping them there. But AI is changing that. Now, with tools to integrate and make sense of data across systems, startups can compete without needing to play by the big guys’ rules. It’s a new era of agility where data orchestration may just be the next big advantage. Exciting times for those rethinking their data strategy! https://lnkd.in/eqGDACYW #AI #DataStrategy #Innovation #Partnership

24

2 Comments

#PET #OWNERSHIP? Reminder of proxy variables Looking at map of pet dog ownership tells… … easily mistaken story. The clear geographical delineation in graph (matching the cold war Iron curtain line) can easily sway you into cultural root cause. 🤦🏻‍♂️ And that would be a great example of the correlation fellacy. Because 25 years after the Iron curtain is down is 2-3 generations of the dogs, long enough time for the political factors to wear off, if they ever had effect. 📖 The real learning of this graph is that there might be hidden factor that is not obvious and still it has profound effect on the actual “target variable”. 🐕 In our dog situation the underlying hidden root cause for the pet ownership is the share of agriculture on total labor market 1.5-2 generations back. Because if you had a farm, you were more likely to get used to living with animals or even needed some (shepherd dogs anybody?). The 2 generations lag is important, because you don’t make pet purchase decisions until you live in your own household and that is roughly, when you start your family (hence when you pass on the generation baton). 🔗 This phenomena is also known as #proxy #parameter and has significant impact also (in positive means) in #MachineLearning. Because proxy parameters are strong predictors, but are insights that markets tend to oversee. So if you know them, proxy parameters give you edge. 💡 Now you know the phenomenon I have a #puzzle for you: It is proven by data in many insurance companies (I worked on one of such studies) that #Zodiac sign is very strong predictor of the #car #accident rate. Yes, really true! Can you guess what it is proxy parameter of?

3

So much fun talking about the power of LLMs with Amazon Bedrock and Anthropic Claude. In summary, the ever-evolving landscape of GenAI, it's becoming clear that there won't be a single LLM (Large Language Model) to rule them all. The rise of application-specific LLMs heralds a new era, where various models will carve out niches to excel in. This diversity opens up a wide playing field for innovation and tailored solutions. When it comes to deploying AI models, striking the right balance in evaluation is paramount. We must weigh factors like cost, speed, and precision to ensure optimal performance and resource utilization. Finding this equilibrium ensures that our GenAI solutions are not just efficient but also effective in real-world scenarios. Anthropic Claude emerges as a frontrunner in the realm of LLMs, distinguished by its exceptional capabilities and groundbreaking advancements. Boasting unparalleled scale, efficiency, and adaptability, Anthropic Claude represents a pinnacle of GenAI innovation. Its ability to seamlessly integrate with Amazon Bedrock amplifies its prowess, providing users with an unparalleled ecosystem for knowledge integration and generation. With state-of-the-art architecture and a commitment to pushing the boundaries of what's possible, Anthropic Claude stands as a testament to the transformative potential of GenAI in shaping the future of technology and beyond. Prompt engineering holds the key to maximizing the potential of applications. Crafting prompts that consider persona, context, tags, and positioning can significantly enhance the model's ability to comprehend and generate relevant responses. This strategic approach empowers AI systems to better serve their intended purposes. In the realm of GenAI architectures, a robust knowledge base is essential. Amazon Bedrock stands out as a key component for architectures like RAG (Retrieval-Augmented Generation), providing a solid foundation to fuel intelligent decision-making and content generation. As we march towards AI integration, bridging the gap with legacy applications is crucial. Agents equipped with open APIs offer a seamless way to facilitate communication and interoperability, ensuring a smooth transition into the AI-driven future. Let's continue to explore, innovate, and collaborate as we navigate the exciting frontier of GenAI together! #aws #LanguageModels #Innovation #TechTrends #AIIntegration #webdevcon #claude #bedrock

68

Snowflake and Databricks are moving aggressively to make more data more accessible by breaking down silos within the enterprise, embracing open data formats and separating data from compute. Meanwhile, an ecosystem of developers in local-first and decentralized tech have been building the primitives to do the same across applications, users, use cases and the web -- a much more revolutionary breakdown of silos. Much more to come on this soon, but for now an earlier primer on data composability and why it matters: https://lnkd.in/dvn9Sx_C

19

📖 Weekend Reading 📖 This week we had profiles on two highly influential figures in tech: Palantir Technologies CEO Alex Karp, and Anduril Industries founder Palmer Luckey. They are in some ways very different individuals, but in others they are radically aligned. 1️⃣ The first piece, on Karp, comes courtesy of Maureen Dowd in The New York Times. Karp appears to have an odd background to be leading a defense-oriented technology company. But what allows the German philosophy doctor to steer Palantir is an extremely principled and unapologetic geopolitical stance: “we have a consistently pro-Western view that the West has a superior way of living and organizing itself.” An interesting observation: “I think we’re in an age when nuclear deterrent is actually less effective because the West is very unlikely to use anything like a nuclear bomb, whereas our adversaries might…Where you have technological parity but moral disparity, the actual disparity is much greater than people think.” Interestingly, while Karp is vocal in his criticisms of some of the Democrats’ policies, he is firmly in the Biden / Harris camp (as a voter and donor) and often butts heads against Palantir’s chairman Peter Thiel when it comes to politics. https://lnkd.in/dMz7BWiV 2️⃣ Next up, Jeremy Stern (Tablet Magazine) profiles Palmer Luckey, another eccentric, highly principled, endlessly fascinating individual. Luckey, who created the Oculus Rift at 19 in a camper trailer, favors anime and collecting decommissioned nuclear missile bases over Karp’s tai chi. But like Karp, he is steadfast in his belief that the West must continue to out-innovate its geopolitical foes (Iran, Russia, China). He also has similar views on rebuilding deterrence, an idea that is core to the company’s mission: “Anduril was founded on the idea the US government had to be reformed so that it could buy deterrence again—and that the only way to deter wars is to be so clearly capable of winning them that no rival power tries to harm core US interests through the use of force.” Anduril is also backed by Peter Thiel, and here, while pursuing what appears like the same objective, Luckey finds himself on the opposite end of the political spectrum as Karp, having famously supported Donald Trump since his first real electoral bid in 2016. I'd recommend this article even just to get a more holistic view of what happened during that time, when Luckey was ousted from Facebook. https://lnkd.in/dpRpnGRF #palantir #ai #llm #defensetech #americandynamism #ukraine #anduril #peterthiel #vr #oculus #fintech #payments #upi

8

2 Comments

It's Friday, Roe AI is on Times Square. We are driving R.O.E for the financial services by bringing the simple, scalable and trustworthy unstructured data intelligence to the data stack. Roe AI enables data teams solve hard problems from documents, enable them to build reliable data pipeline on large documents data https://lnkd.in/eFzzCQz3 Book a time to learn our usecases at getroe.ai

179

3 Comments

I’m headed to Airflow Summit this week! Want to learn more about the state of the art of Airflow data pipeline development and how data engineers everywhere are building better data products with Airflow. If you’re going, want to catch up in SF, or just discuss the joys of healthy and stable data products, let’s meet up. Ps > the image is from a Dalle prompt to “draw me some healthy pipelines”. Everything looks right on the surface but the reality is broken pipelines, leaks, knobs that don’t seem to do anything, and pipes that run on forever. Data engineers: your jobs are safe lol

25

4 Comments

Me and my longtime business partner Uri Merhav are working hard to solve the messy problem of unstructured documents to structured data. Check out Uri's blog post to see him wax poetic about LLMs and the future of document intelligence!

14

1 Comment

There is so much research going into bringing the power of LLMs into the Enterprise context. The main techniques are k-shot learning or prompt engineering, RAG (Retrieval Augmented Generation) & Model Fine tuning (not to mention many variations of these e.g. Agentic RAG, GraphRAG etc.) However, I came across an interesting take on this from IBM and they are calling it "InstructLab" - where the claim is that they are able to teach the LLM "new skills" grounded in an Enterprise context. On the surface it looks like a k-shot learning + fine turning on steroids. Approach seems to be to utilize a small "skills training dataset" in order to generate a large synthetic dataset (perhaps utilizing a specialized LLM) that is then incorporated into the LLM weights natively. This could be a very interesting concept as long as the cost of adding new skills is not very high (when compared to fine tuning). Exciting times! #genAI #enterpriseAI

40

2 Comments

Three years ago, I pitched and received funding for an idea that focused on preventing safety and compliance incidents in target geographies. This was before GenAI was all the craze. We leveraged an ensemble of ML models (ex: CV, NER) that scanned millions of tunnel images in fulfillment centers, and a human-in-loop system that closes the gap when inference confidence was low. Today, I rejoice when I witness other teams working on reducing product damages under Project P.I., which uses a combination of GenAI and Computer Vision to scan 100’s of millions of products, improving the customer experience. I can’t wait to witness what teams across Amazon achieve with GenAI as they focus on solving real-life, papable challenges on behalf of customers. This is the type of GenAI ROI that I believe in! Kudos to them all. You can check more about Project P.I. in the article below: https://lnkd.in/g_SnqhgA cc: Amazon Science #genai #technology #artificialintelligence Set the 🔔 notification on my page, don’t miss a post!

111

13 Comments

Rather than laying off people, maybe Amazon could have loaned those people to teams that need to pay off tech debt, clean up data quality in the store, or tackle whatever items were just below the line in OP1? Just spit balling. Amazon has millions, if not billions, of dollars worth of opportunities just lying around. It also has excess talent. The world's biggest marketplace couldn't figure out how to merge these two things together to add more value for customers and shareholders? Can't spare a two-pizza team to fix recommendations so that after I buy a desk chair it's not recommending me 100 more desk chairs? Improve the excessive time it takes images/variations to load in the mobile app? Fix whatever incorrect SQL query is hiding the fact that delivery promises given at check-out are not being met? On the last one, over the past year more than 50% of my purchases have not been received before the delivery promise initially given. Anecdotally, my friends tell me the same thing. The intriguing thing though, is that once the delivery date is pushed back, I no longer can see the original promise date in Your Orders. I'd wager that somehow this is corrupting the data in the CBR deck and it looks far better than it really is. Either that or 1) The data and anecdotes I have are not representative of the broader population, or 2) Amazon leadership no longer cares that it breaks its promise frequently. I kind of doubt Retail leadership doesn't care about breaking that promise, and I also assume my experience is representative. So that leaves a data issue somewhere. But even if I'm wrong, it's unlikely I'll be renewing my Prime membership when less than 50% of what I order meets the 1-day or 2-day promise.

68

13 Comments

Exciting news for customers: Fine-tuning for Anthropic’s Claude 3 Haiku model is now available in Amazon Bedrock—currently the only managed service to offer this capability. This means our customers can now customize Claude 3 Haiku with their own data to build GenAI applications tailored to their specific needs, while keeping their data secure. Since we launched the preview a few months ago, we've already seen some great use cases. One of our customers, Robin AI, a legal AI company, is using this feature to optimize legal workflows and help lawyers speed up contract processes. With a fine-tuned Claude 3 Haiku in Bedrock, they are helping lawyers finish contract review processes 90% faster. Will be great to see how other customers use this new capability. https://lnkd.in/g2jU8Kfv

324

5 Comments

AI next frontier is possibly developing and scaling the reasoning layer, where AI systems possess hard problem-solving and higher order cognitive intelligence. Came across this good article from Sequoia.

20

1. IMO, agent patterns are one of the most interesting ways to extend LLMs. memary is based on leveraging a graph database (neo4j) for long-term memory (to get around the limited memory inherent in even the largest context windows in LLMs) and augmenting it with other tools like web search and Computer Vision. https://lnkd.in/gfFcBqJx Looks interesting. 2. More about how consolidation effects the entire healthcare industry: https://lnkd.in/gXJKqNJB The Change hack/ransomware attack not only affected claims, service, refills, etc. but now CMS & NCQA are delaying the entire HEDIS reporting process (used for Medicare/Medicaid quality ratings) due to the impact. 3. NARA bans ChatGPT use agency-wide. https://lnkd.in/gW8T94JU Federal Government agencies see this as a risk, including their information leaking and being used to train models (which may then output the wrong information or show PII/PHI other proprietary information). 4. Oh great, now I can't even be angry anymore :) https://lnkd.in/gcrN9mjx

1

1 Comment

Huge Milestone for Data Security: BLOB Scanning for Snowflake is Live at LightBeam.ai. I am thrilled to announce a major leap forward in Data Security for Snowflake users – BLOB scanning is finally here! For those unfamiliar, BLOBs (Binary Large Objects) store all sorts of files within your database, from images and PDFs to audio and more. The challenge? Ensuring sensitive data lurking within these files doesn't slip through the cracks. That's where LightBeam comes in. We seamlessly connect to Snowflake (alongside other data sources) and meticulously scan for PII, PHI, and other sensitive information- even within BLOBs. This empowers you to: - Gain comprehensive data visibility - Identify potential risks associated with structured data - Mitigate security threats swiftly This is a game-changer for data security in the Snowflake ecosystem, and we're pumped to provide this added layer of protection to our customers. Want to experience the power of BLOB scanning for yourself? It's as easy as 1-2-3 Ready to supercharge your data security? Know more: https://bit.ly/4dK1t7q #DataSecurity #Snowflake #BlobScanning

103

4 Comments

What do 4 Ascend.io data 🤓s talk about on a Hackathon hike? 🤖 multi agent orchestration with #LangGraph 🦆 #DuckDB benchmarks and CSV parsing 👷streaming with #Confluent It's an exciting time in data & AI!

60

1 Comment

Distillation vs. Scaling: The Future of AI Lies in Smaller and Smarter Models When it comes to large language models (LLMs), scaling laws tell us one thing: bigger is better for accuracy. But as businesses grapple with the high costs of training, hosting, and maintaining these massive models, we have to ask—is bigger always better? That’s where model distillation comes into play. By transferring the knowledge from a large, sophisticated teacher model to a smaller, more agile student model, we can achieve nearly the same level of accuracy without the enormous resource demands. For the past two years, I’ve been conducting extensive research and experiments on this topic, and I’ve seen firsthand the transformative potential of distillation. #OpenAI just released a pipeline for model distillation—this is a true game changer. This breakthrough enables anyone to distill models efficiently, making cutting-edge AI more accessible and scalable for businesses. The future of AI isn’t just about size—it’s about how we can scale smarter, faster, and with more impact. #AIDistillation #LLMScaling #BusinessAI #RealTimeAI #AIInnovation #LLM #AI #openAI

76

3 Comments

Got It AI's Amol Kelkar explains how an opinionated RAG platform can deliver much Faster Time to Value, Better User Experience, Reliable/Trusted results compared to open source components a team has to string together and work for a while to get an optimized solution. And yet you can have a Choice of LLMs. This is one of the best blogs in #generativeai on RAG. Chandra Khatri Charles H. Martin, PhD Gokhan Tur Dilek Hakkani-Tur Rohan Relan Rohit Kumar Amit Zavery Ashwin Ram Marco Murgia Krishna Hegde Michelle Yee Debanjan Saha Ashish Motivala Saleem Haque Shanmugam Chinnasamy Dipender Saluja Bill Brougher Eros Resmini Gautam Mundewadi Hung Tran Raghav Rastogi Lynne Wander Navin Chaddha Sean Grogan

33

5 Comments

Interesting discussion of the value of proprietary data. Key points: 1 - much of this data doesn’t have value on its own. They cite lack of real marketplaces 2 - it’s still valuabl enough that companies should not share with public llms lest they lost their competitive edge 3 - the use of the data by its owner is aided by clear right to use (first party), inherent utility (already used in some way without ai), and knowledge of space The first point I found the most interesting and controversial. There are many other barriers (privacy, use rights, accessibiiity) that make buying data very hard. Additionally the value of its use in llms is very new. So it may just be a case of the value catching up to overcome the barriers vs no value in it.

3