A great read for anyone building generative AI applications. It touches on actual pain points and the excellent advantages that GenAI applications have. Some highlights include 1. How LinkedIn determined its use case 2. Actual components of the application ( RAG and agents) 3.Automatic evaluation.(Personally this is a favorite of mine) 4.Technical evaluation of latency and throughput . Please read! It’s fascinating. Also found this via Chip Huyen ‘s wonderful post.
Mudra Mukherjee, Ph.D.’s Post
More Relevant Posts
-
If you're in the gen AI space, give this a read. It's accessible yet illuminating, and is a perfect guide to the real challenges of building LLM based applications
Are you building products with generative AI? Karthik Ramgopal and I had a lot of fun developing LinkedIn’s new AI-powered experience, and we’re excited to share both our successes and challenges. We delve into several topics: team composition, agent/system design, evaluation criteria, response quality, cost, latency, and more. Biggest takeaway? Invest early in your evaluation criteria and pipeline. This was a massive effort from an incredible team, and there’s a lot more coming soon. Check out our latest LinkedIn Engineering blog post where we share our experience: https://lnkd.in/gD2Asydd
Musings on Building a Generative AI Product
linkedin.com
To view or add a comment, sign in
-
Experienced Graphic Designer | Branding Specialist | Creative Strategist | Driving Brand Recognition & Engagement | Expertise in Visual Communication & Design | Adobe Creative Suite Digital Communication Specialist
Riding the AI Wave 🌊🏄♀️ Tech evolves rapidly. Keeping up isn't optional, it's vital. The AI game is exploding and reshaping the field. Whether you're into machine learning, data science, AI strategies, or No-Code Tools – the tech world teems with opportunities. Boost your productivity or conquer new tech peaks. The future is here. Don't lag in this tech race! 🚀💻🌐
To view or add a comment, sign in
-
Where to start in AI as a software engineer? Learn the basic concepts of AI models, study different models with different purposes, have a better idea about differences among them, and then start addressing using some models on your local machine or environment. Learn how to develop simple solutions around AI models and grow up the sophistication utilizing advanced concepts such as Agents and RAG. Figure out different use cases and try some such as text, voice, vision, images, and audio. There are piles of potential use cases in each category, and even you may combine multiple categories together. Take your chances and be creative, there are no rules yet in this domain.
To view or add a comment, sign in
-
Does keeping up with the latest in AI make you feel like Sisyphus pushing his boulder? Here’s a decision-making framework I use to decide what to skim and where to go deep: Relevance to My Work: As Head of AI at Amperity, I prioritize advancements that affect: 1. Personalization of customer experiences 2. Workflows of data analysts and engineers 3. Software engineering practices If the problem being solved isn't directly relevant to these areas, I opt for a quick summary of the news. Durability of the Advancement: I focus on techniques that will stand the test of time, and I invest less energy in those that are likely to evolve. For example, passing the right information to LLMs is a problem likely to persist, though specific techniques for solving this problem may shift (fine-tuning, few-shot prompting, RAG, etc.). New Possibilities: I always ask myself, "What can we do now that wasn't possible before? Simple enough, but this framework definitely helps me keep up without getting overwhelmed. Plus, you could always, you know, use AI to summarize the AI news for you… 😉
To view or add a comment, sign in
-
Great post by LinkedIn on ins and outs of building a GenAI product. Many has this perception that LLM can solve all engineering and AI problem. It probably can but with a lot of engineering efforts, as shown in the post. The first time I built a GenAI application I was so excited that I got an “almost there” version immediately, without much effort. Naively I thought I could easily get to the “this is what I want” version in a few days. This turned out to be a continuous effort for a couple of weeks. Just like the post mentioned the first 80% is easy to achieve and to bring it to surpass a 95% requires a much longer time frame. And this is a point that many enterprises (sometimes even developers) do not realise. Just like traditional ML applications, building with LLM also requires technicalities which might look different from what we usually do. Not forgetting the prompting part takes many trials to get to a stable state. If you are building with LLM or Gen AI or just want to know the good and the ugly, go through this post and let me know which points resonated with you! #llm #generativeai #largelanguagemodel #promptengineering
Musings on Building a Generative AI Product
linkedin.com
To view or add a comment, sign in
-
Very insightful article on building products with LLMs. "For product experiences that tolerate such a level of errors, building with generative AI is refreshingly straightforward. But it also creates unattainable expectations, the initial pace created a false sense of ‘almost there,’ which became discouraging as the rate of improvement slowed significantly for each subsequent 1% gain." The key is to knowing which tool (agents, RAG, self-correction, etc.) can help you overcome hurdles and make small progress once you get past the first 80% of the way. https://lnkd.in/eQ6WU48m
Musings on Building a Generative AI Product
linkedin.com
To view or add a comment, sign in
-
Generative AI | Computer Vision | AI | Speech Processing | ADAS & Autonomous Driving | IIM - Ahmedabad
Interesting article on building Generative AI Products. It talks about practical challenges when deploying GenAI in products. Also, insights into unattainable expectations in terms achieving consistent quality and the time required to do so.
Musings on Building a Generative AI Product
linkedin.com
To view or add a comment, sign in
-
🔥 AI Enthusiasts, Your World is About to Get Rocked! 🔥 Listen up, my digital mavens and future-world creators! NVIDIA isn’t just changing the game; they’re handing us the cheat codes with FREE online AI courses. This isn’t your run-of-the-mill "improve your resume" gig. It’s the “hold onto your brains because we’re blasting off to new dimensions” kind of deal. 🤖 Here’s where you redefine your potential: 1️⃣ Generative AI Explained Crack the code on Generative AI and unlock a universe where creation meets technology. Dive in: https://lnkd.in/gTAJ-sKa 2️⃣ Building A Brain in 10 Minutes Unleash the power of neural networks in less time than it takes to grab a coffee. Jumpstart here: https://lnkd.in/gvVrqwZF 3️⃣ Augment your LLM with Retrieval Augmented Generation Redefine intelligence with Retrieval Augmented Generation - the frontier of learning. Explore now: https://lnkd.in/g8hYube9 4️⃣ AI in the Data Center Transform data into decisions, and decisions into dominion. Dive deeper: https://lnkd.in/gvNzawxe 5️⃣ Accelerate Data Science Workflows with Zero Code Changes Make "time-consuming" a term of the past. Accelerate your data science like never before. Get started: https://lnkd.in/gRmxxVn8 🚀 This is Your Launchpad 🚀 Forget playing small. It’s time to wield the tools of tomorrow and craft a future so bright, we’re gonna need shades. This is your ticket to the big leagues, where AI doesn’t just stand for Artificial Intelligence, but for Absolutely Incredible. Share this with someone who’s ready to leap from the edge of innovation into the abyss of greatness. Let’s not just occupy space in the digital era; let’s own it. #aicourses #artificialintelligence #freetraining #dataengineering #businessanalytics
To view or add a comment, sign in
-
-
OCI Gen AI Certified | Software Engineer @ TargetArc | GPA 4.0 | CLSSGB | MS CS @UTA 2024 | Seeking Data Engineer Opportunities
🚀 The One Where Ray is the Hope! 🚀 "Unlocking AI for Cost Reduction with Resource Optimization" Hello everyone! I've been exploring some incredible tools recently, and one that stands out is Ray, an open-source framework that's like a ray of hope for anyone looking to build scalable, distributed applications. Meet Ray, my new best friend in AI and distributed computing! 🌟 What is Ray? 🌟 Think of Ray as a high-performance framework designed to simplify the development of distributed applications. 📈 Why Ray for Cost and Resource Optimization? 📈 From my learner's perspective, Ray is a game-changer because: - Scalability: Seamlessly scale applications across multiple nodes, perfect for handling those massive datasets. - Flexibility: With a unified API, Ray supports various workloads, making it super versatile. - Efficiency: Ray ensures optimal use of computational resources, helping reduce costs and boost performance. ✨ Join Me on This Learning Journey! ✨ Over the next 10 days, I'll be diving deep into how Ray can be leveraged for cost reduction and resource optimization. Each day, I'll share insights, tips, and practical examples—from installation and setup to building and deploying AI models. I'm excited to learn and grow together with all of you. Let's unlock the full potential of Ray and make a real impact. Follow along, ask questions, and let's make this a collaborative journey! #RayOfHope #SelfLearning #AIDeployment #StudentLife #AI #MachineLearning #RayFramework #CostOptimization #ResourceManagement #DataScience #LearningJourney #CostReduction
To view or add a comment, sign in
-
-
🚀 Embarking on a journey in the AI and ML landscape? Discover the 🔑 secrets behind the rapid advancements in Artificial Intelligence and Machine Learning in our latest guide for 2024! Whether you're a beginner or looking to elevate your career, this comprehensive guide is your go-to resource. 🤖✨ Stay ahead in the ever-evolving tech world and explore the myriad opportunities AI and ML offer. Let's shape the future together! 🌐💡 #ArtificialIntelligence2024 #MachineLearningCareers #FutureOfTech #AIInnovation #CareerDevelopment #TechTrends #MLStrategies #EmergingTechnologies #LearnAI #CareerGrowth
To view or add a comment, sign in