Eddie A.

Thank you, Sion Smith, for having us! 🎙 In this talk, Yingjun Wu introduces RisingWave, a scalable streaming database that utilizes #S3 as its primary storage. From underlying technology to real-world scenarios, Yingjun illustrates how RisingWave supports heavy traffic workloads in enterprise-grade production environments. #risingwave #streamingdatabase

3

💥 Python’s not broken—so why are we adding Rust? Two weeks ago, our own Zander Matheson took the stage at PyData NYC (NumFOCUS) to share how Bytewax combines Python’s flexibility with Rust’s power for real-time data processing. 𝐍𝐨𝐰, 𝐭𝐡𝐞 𝐫𝐞𝐜𝐨𝐫𝐝𝐢𝐧𝐠 𝐢𝐬 𝐥𝐢𝐯𝐞! 𝐇𝐢𝐠𝐡𝐥𝐢𝐠𝐡𝐭𝐬 𝐢𝐧𝐜𝐥𝐮𝐝𝐞: - Rust’s role in Bytewax: How it handles the heavy lifting for real-time workflows. - Why PyO3 makes integrating Rust into Python so seamless. - The truth about performance trade-offs and where Rust shines the most. - A live walkthrough of setting up a Python + Rust environment. 🦀 𝐊𝐞𝐲 𝐪𝐮𝐞𝐬𝐭𝐢𝐨𝐧: Is it time to rethink your Python-first strategy for real-time performance? 🐝 Link to the full talk in the comments!

13

‏1‏ ‏תגובה‏

Interesting paper about the limitations of the generative AI. Actually Yann LeCun talks about this limitation as inherited problem in the existing models for long time. Knowing the limitation helps to align expectations. https://lnkd.in/dCid9B_J #generativeai #llm #artificailintelligence #ai

7

‏3‏ ‏תגובות‏

[New on our blog] MLOps Journey: Building a Mature ML Development Process by Albin Sundqvist TL;DR → Building a great AI system takes more than creating one good model. Instead, you have to implement a workflow that enables you to iterate and continuously improve. → Data scientists often lack focus, time, or knowledge about software engineering principles. As a result, poor code quality and reliance on manual workflows are two of the main issues in ML development processes. → Using the following 3 principles helps you build a mature ML development process: 1. Establish a standard repository structure you can use as a scaffold for your projects. 2. Design your scripts, jobs, and pipelines to be idempotent. 3. Treat your pipelines as artifacts, not only the model or data itself. — (link to the full article in the comments) #ml #machinelearning #mlops

2

‏1‏ ‏תגובה‏

🚀 Just slashed our ML feature store latency by 95% - here's how we did it The silent killer in many ML pipelines? Inefficient feature stores. We discovered this the hard way when response times jumped from 50ms to 2 seconds in production. Key optimizations that transformed our performance: Implemented feature computation graphs to eliminate redundant calculations Deployed smart multi-level caching with hot/warm/cold feature strategies Built intelligent batch processing with dynamic sizing The results were dramatic: ✨ Feature computation 📉 CPU utilization 💰 Storage costs 🎯 Model accuracy Check out our detailed technical breakdown and implementation guide here: https://lnkd.in/d3pf6e8P

4

‏1‏ ‏תגובה‏

👉🏼 Evaluating the performance of ChatGPT-4 and ChatGPT-4o in BI-RADS score generation: insights from a radiological analysis 🤓 Dan Nguyen 👇🏻 https://lnkd.in/eC_6VaxN 🔍 Focus on data insights: - 📊 The study tested ChatGPT-4 and ChatGPT-4o on 77 breast cancer images, revealing a combined accuracy of 66.2% for BI-RADS scoring. - 🔍 Performance was markedly higher for BI-RADS 5 cases, achieving scores of 84.4% for ChatGPT-4 and 88.9% for ChatGPT-4o. - ⚖️ Both models faced challenges with lower severity ratings

1

Exploring non-linear relationships in regression? Transformations can lead to trillions of potential models, but is there a better way? Samuele Mazzanti's article discusses how gradient-boosting models can provide effective solutions with ease. #Regression #DataScience

8

‏1‏ ‏תגובה‏

🚀 Check out the latest Medium post by our student Tianyunxi Yin: "Quantization Fundamentals for Model Compression". Explore the essentials of quantization and learn how it helps in compressing machine learning models for better efficiency without compromising performance. Read more here: https://lnkd.in/gHbbR_J4 #USFMDS #DataScience #MachineLearning #ModelCompression #Quantization

15

🧠 What’s a Semantic Cache in RAG? Here’s What We’ve Learned Semantic cache is a time and cost saving method for Retrieval Augmented Generation (RAG). Instead of always running a full vector search, a semantic cache stores embeddings of frequently asked queries and their responses. This speeds up your RAG system by skipping redundant searches for similar queries. How do you implement it? Step 1: Store history of answered questions, their embeddings and the LLM response in a semantic cache table. Step 2: When a new query comes in, compare it against the cache table using vector similarity. Step 3: If there’s a match with score over a certain threshold, serve the cached response. If not, run a fresh retrieval and add it to the cache. Now, let’s talk vector search methods: 1️⃣ Cosine Similarity: Measures angle between vectors, works well for normalized data. 2️⃣ Euclidean Distance: Measures straight-line distance but can be skewed by vector magnitude. 3️⃣ Dot Product: Our favorite—it factors in vector size and performs best for our use cases. Want to see how semantic caching powers a RAG system in real life? Check out our demo at https://lnkd.in/g2jFuHTT. 🚀

2

Open-source #SLMs are closing the gap with large commercial models like #GPT4 💡 Fine-tune those SLMs and they destroy commercial LLMs -> in our benchmarks across 30 tasks fine-tuned Meta Facebook's #llama3.1 outperforms GPT-4 by nearly 20% 😵 🥊 What's even more amazing is that #finetuned open-source SLMs can outperform fine-tuned commercial models like GPT-4o-mini 💰 ➡ See our latest #benchmarks for Llama-3.1 and more on our finding index: https://lnkd.in/gPa-tkrb. ➡ And try out fine-tuning and serving your own SLMs with $25 in free #credits: https://lnkd.in/gTdE8gxh

13

🚀 Looking to streamline fine-tuning for models like Llama 3, Mistral, and Mixtral? Our latest blog shows how Anyscale simplifies LLMs fine-tuning with Ray: 🔑 Key takeaways: • Scalable fine-tuning with LLM-Forge • Model Deployment with RayLLM Read more: https://lnkd.in/gQwD4A2U

35

A great read on the performance and financials of LLM fine-tuning. Really hoping the outcome is a tool that helps overcome the biggest elephant in the room, Mr. Cost! #llm #finetuning #finance #computaioncosts

2

‏1‏ ‏תגובה‏

Kolmogorov–Arnold Networks seem like a paradigm shift in ML architectures. Curious to see how they evolve. "While MLPs have fixed activation functions on nodes ("neurons"), KANs have learnable activation functions on edges ("weights"). KANs have no linear weights at all -- every weight parameter is replaced by a univariate function parametrized as a spline. " https://lnkd.in/gdPk_qCQ

29

Apache Kafka and Confluent power a real-time prediction model that helps prevent the risk of deadly sepsis in bone marrow transplant patients. #DataScience #AI by Cameron Carlin and Sergei Semenov thanks to Confluent

1

SLASH your RAG memory usage by 32X and speed up document retrieval (Without compromising accuracy significantly) 👇 If you are dealing with: - Massive datasets - Expanding data scale - Sluggish retrieval times - Escalating compute costs I am having my share, too. ↳ Projects scaling up ↳ Data piling up ↳ Memory usage climbing ↳ Retrieval speed dropping ↳ Costs shooting up We all are in the same boat. But I found a solution that actually works. 𝗩𝗲𝗰𝘁𝗼𝗿 𝗤𝘂𝗮𝗻𝘁𝗶𝘇𝗮𝘁𝗶𝗼𝗻 (𝗩𝗤). Here’s how VQ can help you: ✅ Cuts down memory needs ✅ Reduces latency ✅ Slashes cost Weaviate offers VQ solutions in 2 ways: 1/ 𝗣𝗿𝗼𝗱𝘂𝗰𝘁 𝗤𝘂𝗮𝗻𝘁𝗶𝘇𝗮𝘁𝗶𝗼𝗻 (𝗣𝗤) 𝗪𝗵𝗮𝘁 𝗶𝘁 𝗗𝗼𝗲𝘀: Compresses your vector embeddings by breaking them down into smaller, manageable segments. 𝗕𝗲𝗻𝗲𝗳𝗶𝘁𝘀: Reduces your storage needs by almost 24 times while balancing performance and recall. 𝗕𝗲𝘀𝘁 𝗙𝗼𝗿: Those who use HNSW indexes and need a fine balance between speed and accuracy. 2/ 𝗕𝗶𝗻𝗮𝗿𝘆 𝗤𝘂𝗮𝗻𝘁𝗶𝘇𝗮𝘁𝗶𝗼𝗻 (𝗕𝗤) 𝗪𝗵𝗮𝘁 𝗶𝘁 𝗗𝗼𝗲𝘀: Converts each vector into a binary format, drastically reducing the size from bytes to bits. 𝗕𝗲𝗻𝗲𝗳𝗶𝘁𝘀: Achieves a 32x reduction in storage requirements and speeds up search processes. 𝗕𝗲𝘀𝘁 𝗙𝗼𝗿: Projects where speed is critical and slight compromises on accuracy are acceptable. 📚 𝗧𝗿𝗮𝗱𝗲-𝗼𝗳𝗳𝘀: → PQ might slightly reduce recall but saves more memory. → BQ offers incredible speed at the cost of some accuracy. PQ might work for you. Or BQ. It really depends on your project-specific needs and constraints. 📚 Want to learn more about VQ (and Weaviate)? Follow: 👉 Philip Vollet 👉 Leonie Monigatti 👉Connor Shorten 👉Victoria Slocum 👉 Erika Cardenas 𝗖𝗵𝗲𝗰𝗸 𝗰𝗼𝗺𝗺𝗲𝗻𝘁𝘀 𝗳𝗼𝗿 𝗺𝗼𝗿𝗲 𝗿𝗲𝘀𝗼𝘂𝗿𝗰𝗲𝘀👇 --------------- 📌 If you enjoyed this (and want to support us): → 𝗟𝗶𝗸𝗲 👍 → 𝗥𝗲𝗽𝗼𝘀𝘁 ♻️ Thanks! 📌 𝗣.𝗦. 𝗝𝗼𝗶𝗻 𝗴𝗿𝗼𝘄𝗶𝗻𝗴 5000+ 𝗤𝗲𝗻𝗱𝗲𝗹 𝗔𝗜 𝗰𝗼𝗺𝗺𝘂𝗻𝗶𝘁𝘆

73

‏5‏ ‏תגובות‏