Sameer Paranjpye

So nice to see Bratin Saha, AWS VP/GM, at the center stage joining our Co-Founder and CEO at the Imagine Keynote today! We are a Strategic Partner for AWS and have been steadily building upon our multi-year relationship. Today we are excited to announce our integration with Amazon Q in the conversation automation space. Amazon Q for Business Users is a Gen AI powered assistant that can be used seamlessly within Automation Anywhere's A360 platform by our customers to answer questions, get summaries and generate content and more importantly complete tasks within the enterprise context.   Here is what Mihir Shukla, our Co-Founder and CEO said about the power of this integration: “By making conversational automation available enterprise-wide with AWS Q, we’re arming every employee in companies with the potential to automate and transform their challenging process workflows to transform the way they work and reduce the time it takes from days and weeks to minutes”. Find out more in this press release: https://lnkd.in/g23rmVYa Prince Kohli, Adi Kuruganti, Tom Garland, Jay Bala, Peter White, Steve Shah, Claudia Michon, Matt Wood, Jagjit Dhaliwal (JD), Omar Zarka, Olawale Oladehin, Asheesh Mangla, Imran Moin, Watson G., Tanner McRae, Pranjal Raje, Dayanand Thakur, Gabe Akisanmi, Kartik Dalal, ShreeLakshmi Krishnamurthi, Charu Venkatraman, Sarangadhar Sahani, Geoffrey LAISSUS, Shane Patel #AAImagine, #AWS, #AmazonQ, Automation Anywhere, Amazon, #AI, #GenerativeAI

104

3 Comments

🎉 Today marks a very special milestone in our journey with #ApachePulsar – we’re celebrating 6 years as a top-level Apache Software Foundation project! 🚀 It’s been an incredible experience watching Pulsar evolve from an idea into one of the most unique and advanced messaging and data streaming platforms out there. What started as a vision to build a system that could handle the complexity of real-time data at scale has grown into something beyond what I could have imagined. The innovation and dedication of the community, the contributors, and everyone who believed in Pulsar's potential have been the true driving force behind its success. From its unique multi-tenancy, geo-replication, and no-data-rebalance capabilities to its ability to power everything from small startups to large enterprises, Pulsar is a testament to what can be achieved through collaboration and open source. Looking back on the journey, I’m beyond proud of what we’ve accomplished, and excited about what’s next. With Pulsar 4.0, #Ursa Engine, UniConn (Universal Connectivity), and even bigger ideas on the horizon, the future is brighter than ever. 🌟 Thank you to everyone who’s been a part of this ride, and here’s to many more years of innovation and community building! 🙌 Check out the latest blog from Matteo Merli and me reflecting on this incredible milestone: https://lnkd.in/gV5W8rQy #ApachePulsar #OpenSource #DataStreaming #Messaging #Queuing #Community #Anniversary #Ursa #UniConn

105

5 Comments

🔍 Unlocking the secret sauce to efficient AI models! 🧩 Puzzle by #NVIDIAResearch isn't your average jigsaw puzzle - it's a mind-bending framework transforming over-parameterized LLMs into hardware-optimized masterpieces! 🌟 Let's put the pieces together for smarter, faster, and greener AI magic! 🚀 #AI #Innovation #PuzzleSolvers 🧠✨ #innovation

3

The speed at which these models improve is staggering, so is the cost to build them. Only way to keep up is to leverage principles such as AWS Bedrock so you have the flexibility you need to quickly select the model that works best for your task. https://lnkd.in/gNAR63YC https://lnkd.in/g_FtP6e9

36

In the latest episode of Prodcast, Silvia Botros (SRE Architect, Twilio, and author of "High Performance MySQL, 4th edition") and Niall Murphy (Co-founder & CEO, Stanza) join hosts Steve McGhee and Jordan Greenberg, to discuss many topics near and dear to reliability: cultural shifts in database engineering, rate limiting, load shedding, holistic approaches to reliability, proactive measures to build customer trust, and much more!

1

This has been an eventful week for the entire data streaming ecosystem from StreamNative, Confluent, and Redpanda Data. There's no doubt that data streaming is undergoing rapid and transformative developments. It’s particularly interesting to see RedPanda embracing our "New CAP Theorem" (https://lnkd.in/gcuqViBm) by offering cloud topics and following our lead, as well as Confluent’s, in introducing Iceberg integration. We’re excited to see our peers following the path set by the Ursa engine, as it represents what a future-proof data streaming platform should look like. https://lnkd.in/gRZZWC_7 🔺 Tackling the New CAP Theorem: It must offer flexibility in balancing Cost, Availability, and Performance. At StreamNative, we address this with two WAL backends—latency-optimized disk-based replication (via BookKeeper) and cost-optimized object storage replication (via S3). ☁ Flexible Deployment Options: It must be deployable across a range of environments—from Public Cloud to Private Cloud—with options for BYOC, Dedicated, and Serverless deployments. This gives customers the flexibility to choose the deployment that best suits their needs without compromising feature parity, performance, security, or sovereignty. https://lnkd.in/gVQr8867 🌊 Multi-Modal Engine: It must be capable of handling table-stream duality. Streaming data directly into a lakehouse is becoming a must-have feature. Reducing barriers between data streams and lakehouses accelerates data usage and enables enterprise-ready, real-time data for GenAI applications. 🔗 Multi-Semantics via Multiple Protocols: It should support both queuing and streaming semantics, allowing customers to choose the right option for their workloads—whether it's low-latency queuing for transactional use cases or high-volume data streaming for less latency-sensitive applications. At StreamNative, we enable this through support for multiple protocols, including Pulsar, Kafka, MQTT, and others. 👯 Multi-Tenancy: Multi-tenancy must be built in from the start, preventing teams from having to manage multiple small clusters within a single organization. #ApachePulsar stands out as the leader in multi-tenancy, being the only data streaming solution that natively supports this feature. It’s no surprise that more engines will start to look like Ursa in the next 2-3 years. If you're attending the Current Summit next week, stop by our booth. Let’s talk about Ursa, Kafka, and the future of data streaming!

55

1 Comment

Kafka is reborn cloud-native! Confluent acquired Warpstream. But Confluent also has Kora. Which one will be the successor, and what’s next? History has some lessons. I honestly did not expect this so soon. Confluent still moves as fast as Kafka. For those of you who don’t know Warpstream, you’ll know it soon. it’s cloud-native Kafka made simple. Just bring your own cloud (BYOC) and drop Warpstream in. Its pricing and simplicity made Kafka REALLY cost-effective. It’s a bold, great acquisition by Confluent. What’s next? If history tells us anything, and it does, here’s what we might expect to happen. 1. Kora will “evolve” into Warpstream. And that’s OK. Confluent now has two cloud-native Kafkas. Confluent has already made it clear that Warpstream is their new Kafka BYOC solution. That’s usually acquisition-speak for Warpstream is becoming part of the new foundation. So what happens to Kora? It probably evolves into managed Warpstream behind the existing UI and API. This isn’t the first time Confluent has built something and then acquired a better alternative. Think Flink. This change really shouldn’t matter much for existing customers because it will be hidden. 2. Bento must die. Warpstream committed to keeping a version of the original Benthos alive and open source. Will Confluent back it now? Is that part of the strategy? This one is a little harder. My guess is standalone Bento is done. There aren’t a lot of Warpstream customers. But many people depend on Connect and Flink. Let’s hope Flink gets a Bento box. 3. New messaging protocols will continue to emerge. The Warpstream acquisition adds fuel to the fire. People want lower cost Kafka, and “Confluent BYOC” will most likely become more expensive. Higher costs will continue to drive companies to look for alternatives for messaging and integration. This is good for other vendors: - RedPanda and other Kafka-compatible solutions will continue to grow - Estuary Flow is built on Gazette, which is cloud-native exactly-once delivery with stream store and replay. It already supports the Kafka consumer API for compatibility. - IoT vendors like NATS continue to grow. It’s a different market. - We’re also long overdue for new messaging vendors to replace JMS, RabbitMQ and others. 4. Kafka will become hidden plumbing. The question is, for what? The acquisition doesn’t change this at all. Data integration, real-time analytics, AI, and other technologies will hide Kafka and other messaging protocols. Estuary Flow hides Gazette. ELT vendors that use Kafka hide it. Even Confluent is innovating on top of Kafka. What do you see coming next? What will Confluent do next? Estuary Confluent WarpStream Apache Kafka Redpanda Data Synadia

110

25 Comments

Folks I've spoken to about Amazon have likely heard me talk about the importance of the PR/FAQ process, a.k.a. "the working backwards process". This blog posts gives a very nice example of a PR/FAQ. https://lnkd.in/edTeuVRQ. Note that the blog doesn't capture the presumably 100s of reviews/discussions/rewrites that this doc represents, which is a part of the real value. When the process is done it all looks so simple, but getting a large organization to make a decision to something simple and do it in a BIG way is the challenge the fun of it all.

63

2 Comments

Today, Sonatype was named the Leader in the latest Forrester Wave on Software Composition Analysis (SCA). Not only did we land in the top spot, but Forrester gave Sonatype the highest scores for component detection and analysis, SBOM generation and ingestion, and more. For me, open source malware is one of the biggest threats that continues to be missed by traditional SCA and endpoint security solutions. This is something we’ve spoken with customers about for years. I was especially pleased to see Forrester’s focus on the importance of malicious package detection — reference customers told Forrester that Sonatype is an “excellent choice” to manage malicious package risk. Organizations today are facing increasing regulatory pressure, mounting cybersecurity threats, and persistent technical debt, all while shipping innovative, transformative software at breakneck speed. Our mission is to help businesses, developers and security teams face these threats head-on without sacrificing development speed. Check out the full Forrester report: https://lnkd.in/eN_PHGZS #SoftwareSupplyChain #OpenSource #SoftwareCompositionAnalysis #OpenSourceSecurity

184

13 Comments

This is a great first step in understanding how scaling dynamics extend beyond language to other types of sequential data. Time is a universal axis, understanding how models learn optimally over data collected over time is a crucial step in the broader AI research agenda.

26

What keeps Postgres thriving after three decades? The strength of its open source community, the spirit of innovation, and its ability to evolve with each new data challenge. In our latest blog, EDB Vice President and Postgres Evangelist Bruce Momjian reflects on the Postgres journey and how its extensibility continues to shape the future of data. Postgres isn’t just a database; it’s a way of life. From its vibrant community of contributors to the continuous advancements in data handling, Postgres embodies the power of collaboration. If you’re passionate about technology and want to make an impact, then the Postgres community is where you want to be. Read the blog and see how you can be part of this journey ➡️ https://lnkd.in/e3shY6yw #Postgres #PostgreSQL #justsolveitwithPostgres #EDBPostgresAI #data #AI #database #DBA #opensource

6

Excited to be open-Sourcing long context (128K) Llama-3 70B - Llama-3-Giraffe-70B and Giraffe-Instruct From Abacus.AI The biggest issue with the Llama-3 family is that they are fairly small context models. So, even though they are as performant as GPT-4 on critical benchmarks, they can only be used in the real world with a bigger context. We just open-sourced two 128K long-context fine-tunes for Llama-3-70B, which effectively puts it on par with GPT-4 regarding context length support and real-world usage. At Abacus AI, we already use Llama-3 on a few production workloads as a great base model for our customer fine-tunes. Links to both models - https://lnkd.in/gPX_XpD6 https://lnkd.in/gaeEq94t

3

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

With Qwak's MLOps solution integrated into JFrog's platform, customers can manage their AI models with JFrog Artifactory serving as the model registry and JFrog Xray ensuring the security of ML models

2

Scale at ease, Lower cost & Future proof

2

Graviton is a super efficient (and green) way to manage a lot of the heavy lifting around GenAI and AI/ML... both in prepping the training data and doing the inference.

15

We have written this blog to discuss the differences in model quality across AI cloud providers on the recent Meta's Llama3.1 models (405B, 70B and 8B). In other words, same model served by different providers can come out with very different results. At #Together, we believe that inference services are not fungible. Differences in implementation decisions, optimizations, and quality testing processes can mean the difference of a percentage point or more on a benchmark, and a success or a failure for your application task. We provide our five-pillar of quality testing recipe to ensure high model serving quality in a transparent way. #quality #genai #inference #LLM #Llama3.1

22