TwoSetAI
TwoSetAI
  • 52
  • 255 173
What Happens When You Combine RAG with Text2SQL?
🤔 Looking to improve your Text2SQL performance?
In this episode, join Angelina and Mehdi, for a discussion about a real world industry use case of RAG with Text2SQL systems.
Who's Angelina: VP of AI and data, Co-founder of Transform AI Studio, two-time fast.ai fellows under Jeremy Howard, published author. www.linkedin.com/in/MeetAngelina/
Who's Mehdi: Professor of Computer Science, Co-founder and Chief AI Engineer at Transform AI Studio, NSF fellow, published author. www.linkedin.com/in/mehdiallahyari/
What You'll Learn:
🔎 Explain what QueryGPT is and why Uber developed it
🚀 Highlight the challenges Uber faced with data queries
🛠 Walk through of the evolution of QueryGPT's architecture (20 iterations!)
🎯Walk away with key learnings and implications for other companies (build vs. buy)
✏️ In This Episode:
00:00 Intro
03:08 Why QueryGPT?
03:58 Initial design and architecture
06:48 Key challenges
07:36 Latest architecture
08:26 AI Agents used in the system
13:23 Continuous evaluation
14:25 Key learnings from QueryGPT
15:35 Fine-tuning
🦄 Any specific contents you wish to learn from us? Sign up here: noteforms.com/forms/twosetai-youtube-content-sqezrz
🧰 Our video editing tool is this one!: get.descript.com/nf5cum9nj1m8
🖼️ Blogpost for today: www.uber.com/en-IN/blog/query-gpt/?uclick_id=6cfc9a34-aa3e-4140-9e8e-34e867b80b2b
📬 Don't miss out on the latest updates - Subscribe to our newsletter: mlnotes.substack.com/
📚 If you'd like to learn more about RAG systems, check out our book on the RAG system: angelinamagr.gumroad.com/
🕴️ Our consulting firm: We help companies that don't want to miss the boat of the current wave of AI advancement by integrating these solutions into their business operations and products. www.transformaistudio.com/
Stay tuned for more content! 🎥 Thanks you for watching! 🙌
มุมมอง: 6 677

วีดีโอ

Can AI Agents Revolutionize How We Work With Excel Data?
มุมมอง 12Kวันที่ผ่านมา
🤔 Looking to combine the power of AI Agents and Text2SQL? In this episode, join Angelina and Mehdi, for a discussion about Agentic Text2SQL. Who's Angelina: VP of AI and data, Co-founder of Transform AI Studio, two-time fast.ai fellows under Jeremy Howard, published author. www.linkedin.com/in/MeetAngelina/ Who's Mehdi: Professor of Computer Science, Co-founder and Chief AI Engineer at Transfor...
Slack's GTM Expert Reveals the SECRET to Hypergrowth!
มุมมอง 22914 วันที่ผ่านมา
🤔 Looking to go to market? Join this insightful episode where Holly Chen, who scaled Slack from $100M to $700M and took it public. She is now an interim CMO and growth advisor, and she shared with me some of her expert marketing strategies for startup growth. Don't miss Holly's expertise on building traction and making the right plays in the competitive tech industry. ✏️ In This Episode: 00:00 ...
How Contextual Retrieval Elevates Your RAG to the Next Level
มุมมอง 37K21 วันที่ผ่านมา
🤔 Looking to enhance your RAG performance? Before we dive in, we have some exciting news! Our RAG live course is coming up soon, and as a way of giving back to our amazing community, we're offering you 15% off. Just use this link: maven.com/angelina-yang/mastering-rag-systems-a-hands-on-guide-to-production-ready-ai?promoCode=TwoSetAI We'd love to see you there! 🎉 In the course, you'll have the ...
Agentic RAG Explained - Build Your Own AI Agent System from scratch! (Step-by-step code)
มุมมอง 43Kหลายเดือนก่อน
🤔 Looking for using AI Agents with RAG? In this episode, join Angelina and Mehdi, for a discussion about Agentic RAG. Who's Angelina: VP of AI and data, Co-founder of Transform AI Studio, two-time fast.ai fellows under Jeremy Howard, published author. www.linkedin.com/in/MeetAngelina/ Who's Mehdi: Professor of Computer Science, Co-founder and Chief AI Engineer at Transform AI Studio, NSF fellow...
How to become more visible in your field?
มุมมอง 6Kหลายเดือนก่อน
🤔 Looking to establish strong online presence or build personal brand? In this episode, join Angelina and Mehdi, for a quick intro of Oscr AI - an AI toolkit that boosts your brand's reach. Website: www.oscr.tech/index.html Join our discord here: discord.com/invite/qQ2a4nKRt2 Who's Angelina: VP of AI and data, Co-founder of Oscr AI and Transform AI Studio, two-time fast.ai fellows under Jeremy ...
Production RAG Secrets the Pros Don't Want You to Know -- Part 2
มุมมอง 552หลายเดือนก่อน
🤔 Looking for the ultimate roadmap for implementing your RAG in production? In this episode, join Angelina and Mehdi, for a discussion of recommended approaches for each step of building your RAG system in production. Who's Angelina: VP of AI and data, Co-founder of Transform AI Studio, two-time fast.ai fellows under Jeremy Howard, published author. www.linkedin.com/in/MeetAngelina/ Who's Mehdi...
Best Practices for Building Production RAG - Part 1
มุมมอง 15Kหลายเดือนก่อน
🤔 Looking for the ultimate roadmap for implementing your RAG in production? In this episode, join Angelina and Mehdi, for a discussion of recommended approaches for each step of building your RAG system in production. Who's Angelina: VP of AI and data, Co-founder of Transform AI Studio, two-time fast.ai fellows under Jeremy Howard, published author. www.linkedin.com/in/MeetAngelina/ Who's Mehdi...
Frustrated with Ambiguous User Queries? You Can't Miss Query Expansion! (Step-by-step code demo)
มุมมอง 4982 หลายเดือนก่อน
🤔 Looking to improve your search retrieval when user queries are vague? In this episode, join Angelina and Mehdi, for a discussion about "Query Expansion" for your RAG systems. Who's Angelina: VP of AI and data, Co-founder of Transform AI Studio, two-time fast.ai fellows under Jeremy Howard, published author. www.linkedin.com/in/MeetAngelina/ Who's Mehdi: Professor of Computer Science, Co-found...
The BEST Mental Model for Optimizing Your LLMs - Part 2
มุมมอง 16K2 หลายเดือนก่อน
🤔 Looking to optimize LLM-based RAG system, but don't know where to begin? In this episode, join Angelina and Mehdi, for a follow up discussion about strategies to optimize your LLM performance suggested by OpenAI. If you haven't checked out Part 1 of this video, click here: th-cam.com/video/Ob4KePkOI8Q/w-d-xo.htmlsi=nPYO92uU-a1n6FNa Who's Angelina: VP of AI and data, Co-founder of Transform AI...
The BEST Mental Model for Optimizing Your LLMs - Part 1
มุมมอง 15K2 หลายเดือนก่อน
🤔 Looking to implement your own LLM-based RAG system, but don't know where to begin? In this episode, join Angelina and Mehdi, for a discussion about strategies to optimize your LLM performance. Part 2 is ready! check here: th-cam.com/video/zZJ7UFWL-yU/w-d-xo.html Who's Angelina: VP of AI and data, Co-founder of Transform AI Studio, two-time fast.ai fellows under Jeremy Howard, published author...
Bridging Fashion & AI: Insights with Summer Wang, Fashion Entrepreneur
มุมมอง 16K2 หลายเดือนก่อน
🤔 Looking to build AI for fashion industry? In this episode, we welcome Summer Wang, a serial entrepreneur in the fashion industry and co-founder of Eukari, a New York-based fashion brand. Summer shares her journey from growing up in an industrial district in China to becoming a fashion entrepreneur. She discusses integrating AI into her business, the current capabilities and limitations of AI ...
Generative Feedback Loops: The Secret Weapon for Supercharging Your RAG System
มุมมอง 16K2 หลายเดือนก่อน
🤔 Looking to enhance your RAG systems and save costs? In this episode, join Angelina and Mehdi as they explore Generative Feedback Loops. Discover how this innovative technique can improve your AI's performance, reduce bias, and create a self-evolving knowledge base. Who's Angelina: VP of AI and data, Co-founder of Transform AI Studio, two-time fast.ai fellows under Jeremy Howard, published aut...
SearchGPT is Google? Fireside Chat with AI Researcher
มุมมอง 5102 หลายเดือนก่อน
🤔 Looking for honest reviews about SearchGPT from AI researcher and industry expert? In this episode, join Angelina and Mehdi, for a fireside chat about SearchGPT. This new tool promises to revolutionize the way we find and interact with information online, blending ChatGPT with real-time access to the internet. Who's Angelina: VP of AI and data, Co-founder of Transform AI Studio, two-time fast...
Best in Class Image Extraction with Fine-Tuned Vision Language Model
มุมมอง 7663 หลายเดือนก่อน
🤔 Looking to extract images out of PDFs? In this episode, join Angelina and Mehdi, for an intro of a cool fine-tuned vision language model that can accurately capture images, figures and tables in academic papers . Who's Angelina: www.linkedin.com/in/MeetAngelina/ Who's Mehdi: www.linkedin.com/in/mehdiallahyari/ What You'll Learn: 🔎 The very practical python library that you can use to achieve ...
GraphRAG - Will the Dust Settle?
มุมมอง 2.5K3 หลายเดือนก่อน
GraphRAG - Will the Dust Settle?
HippoRAG - Revolutionizing AI Retrieval with a 20-Year-Old Algorithm!
มุมมอง 2.1K3 หลายเดือนก่อน
HippoRAG - Revolutionizing AI Retrieval with a 20-Year-Old Algorithm!
Step-by-step Code for Knowledge Graph Construction
มุมมอง 4.8K4 หลายเดือนก่อน
Step-by-step Code for Knowledge Graph Construction
The Secret Weapons Revolutionizing Customer Support: RAG + Knowledge Graphs
มุมมอง 1.4K4 หลายเดือนก่อน
The Secret Weapons Revolutionizing Customer Support: RAG Knowledge Graphs
Get Ahead in Your RAG! Must-Know Knowledge Graph Integration Secrets!
มุมมอง 2.1K4 หลายเดือนก่อน
Get Ahead in Your RAG! Must-Know Knowledge Graph Integration Secrets!
Crack the Nut of Advanced RAG - The Power of Knowledge Graph
มุมมอง 3K5 หลายเดือนก่อน
Crack the Nut of Advanced RAG - The Power of Knowledge Graph
Rag Architecture and Data Ingestion Pipeline
มุมมอง 3035 หลายเดือนก่อน
Rag Architecture and Data Ingestion Pipeline
AI for Everyone: How to Leverage AI Without Tech Skills
มุมมอง 3305 หลายเดือนก่อน
AI for Everyone: How to Leverage AI Without Tech Skills
RAG for Beginners! Step-by-Step Tutorial Using Jupyter Notebook
มุมมอง 2.3K5 หลายเดือนก่อน
RAG for Beginners! Step-by-Step Tutorial Using Jupyter Notebook
Meet CRAG: Don't Miss This Simple Hack for your Production RAG!
มุมมอง 1.1K5 หลายเดือนก่อน
Meet CRAG: Don't Miss This Simple Hack for your Production RAG!
We Won Anthropic's Developer Contest with Our STORM Implementation! (Step-by-step Code Walkthrough)
มุมมอง 6726 หลายเดือนก่อน
We Won Anthropic's Developer Contest with Our STORM Implementation! (Step-by-step Code Walkthrough)
Don't Deploy RAG Without This! Embedding Quantization (~4 lines of code)
มุมมอง 6876 หลายเดือนก่อน
Don't Deploy RAG Without This! Embedding Quantization (~4 lines of code)
Human vs. Machine: Crafting Long-Form Articles with AI Ghostwriters
มุมมอง 6546 หลายเดือนก่อน
Human vs. Machine: Crafting Long-Form Articles with AI Ghostwriters
Privacy-aware -The FREE Desktop App Taking On ChatGPT!
มุมมอง 5866 หลายเดือนก่อน
Privacy-aware -The FREE Desktop App Taking On ChatGPT!
RAG Production Trick - Semantic Cache (Step-by-step Juicy Code Walk-Through)
มุมมอง 1.9K6 หลายเดือนก่อน
RAG Production Trick - Semantic Cache (Step-by-step Juicy Code Walk-Through)

ความคิดเห็น

  • @paulmiller591
    @paulmiller591 6 ชั่วโมงที่ผ่านมา

    Thanks guys very useful

  • @vijaygharge2414
    @vijaygharge2414 13 ชั่วโมงที่ผ่านมา

    Quite informative!

  • @wftrdshometoprofessionalfo142
    @wftrdshometoprofessionalfo142 3 วันที่ผ่านมา

    Special thanks to Mehdi, it is pleasure if we could find a field for cooperation. Kindly let us know how to communicate to you for having discussions on it. Regards

    • @TwoSetAI
      @TwoSetAI 3 วันที่ผ่านมา

      you can reach out to us here: Angelina@oscr.tech. Thanks

  • @123456crapface
    @123456crapface 3 วันที่ผ่านมา

    I guess anyone is an AI guru these days 😂

    • @TwoSetAI
      @TwoSetAI 3 วันที่ผ่านมา

      @@123456crapface anyone willing to try and use AI can be a guru. Especially with more and more low code no code tools. Anyone can be enabled

  • @DaviZucon
    @DaviZucon 4 วันที่ผ่านมา

    Well done ! I would like to see a comparison in terms quality and scale for classification between a in house trained models vs LLMs !

    • @MehdiAllahyari
      @MehdiAllahyari 4 วันที่ผ่านมา

      Here's a great blog post that hopefully answers your question. They have compared the results of an LLM (Llama-3.1-8B) with a small model. They demonstrate that small trained classifier outperforms LLM especially in few-shot learning. Here's the link: huggingface.co/blog/sdiazlor/custom-text-classifier-ai-human-feedback But in general, scaling an LLM for classification is hard, dealing with latency, cost, etc in general is challenging.

    • @MehdiAllahyari
      @MehdiAllahyari 3 วันที่ผ่านมา

      I rewrite the comment just in case the previous one is not published. Here's a blog that compares LLM(i.e. Llama 3.1) and a small trained classifier. And the trained classifier outperform LLM model. Here's the link: huggingface.co/blog/sdiazlor/custom-text-classifier-ai-human-feedback

    • @MehdiAllahyari
      @MehdiAllahyari 2 วันที่ผ่านมา

      Take a look at this blog post. They have done the very same thing, and compared an LLM (i.e. LLama3.1) with a small trained model and demonstrate that trained model outperforms the LLM. Here's the link: huggingface.co/blog/sdiazlor/custom-text-classifier-ai-human-feedback

    • @MehdiAllahyari
      @MehdiAllahyari 13 ชั่วโมงที่ผ่านมา

      For some reason, my comment doesn't show up here since it has a link. Search for this "How to build a custom text classifier without days of human labeling", it's blog that has compared llm(Llama3.1) with a small trained model and they show that the small model actually outperform the llm.

  • @ChiTien-Hsieh
    @ChiTien-Hsieh 4 วันที่ผ่านมา

    Great talk! One small point I’d like to mention is that at around 17:55, Angelina “hmm”s five times within the next 15 seconds, which is quite distracting. While this habit might work well in an offline meeting where such sounds signal active listening, in an online setting, it can actually interrupt the flow and impact the quality of the talk-especially when I’m trying to focus on Mehdi’s insights. A little nodding or some sign language with the mic muted would be really appreciated! Anyway, it was a very insightful talk-I’m just nitpicking.

    • @TwoSetAI
      @TwoSetAI 4 วันที่ผ่านมา

      Thank you for your feedback!

  • @abbyning8498
    @abbyning8498 5 วันที่ผ่านมา

    What's new in this compared to Anthropic's post?

  • @Mars.2024
    @Mars.2024 5 วันที่ผ่านมา

    Hi there ; I feel a quick rewrite or just a short conversation between llm and user to clarify what the user meant is a much better and cheaper approach.

    • @TwoSetAI
      @TwoSetAI 3 วันที่ผ่านมา

      thats a potential solution! paste a link to your project here if you want to share your findings with everybody else. Thanks!

  • @Mars.2024
    @Mars.2024 5 วันที่ผ่านมา

    Thank you for this awesome video 🌱🧝‍♀️

  • @limjuroy7078
    @limjuroy7078 5 วันที่ผ่านมา

    Great job! Keep up the excellent work!

  • @gamekhela
    @gamekhela 6 วันที่ผ่านมา

    How to chunk it, can you share how to do with local LLM and SQL server database

  • @RedCloudServices
    @RedCloudServices 11 วันที่ผ่านมา

    Another solution to the similarity issues is not to use a cache at all

    • @MehdiAllahyari
      @MehdiAllahyari 11 วันที่ผ่านมา

      True if you're not implementing it at scale. Otherwise, cache is a critical component. :)

  • @SiaBarz
    @SiaBarz 11 วันที่ผ่านมา

    Dorud Mehdi jan, awesome as always

    • @MehdiAllahyari
      @MehdiAllahyari 11 วันที่ผ่านมา

      Mamnoon Metalika jan!

  • @SabZuso
    @SabZuso 12 วันที่ผ่านมา

    I think there is one small problem about querying slightly different values. What if user enters values that are slightly different? For example, instead of "Warner Bros", user might enter "warner bros", or "Warner brothers", or "WB". Also, there are different spellings and foreign names for movies, or some movies have "the" such as "The Truman Show" vs. "Trueman Show".

    • @MehdiAllahyari
      @MehdiAllahyari 12 วันที่ผ่านมา

      You are absolutely right. Since i didn't preprocess/clean the text, if you give "WB" for example, it wouldn't work. My implementation is the basic version and there is a lot of room for improvement. :)

    • @SabZuso
      @SabZuso 11 วันที่ผ่านมา

      @@MehdiAllahyari Thank you! Do you have any thoughts how to handle this situation? I could not find a solution for this problem.

    • @MehdiAllahyari
      @MehdiAllahyari 11 วันที่ผ่านมา

      @@SabZuso The best solution I can think of is to store unique values of columns like "studio" (because number of unique values is not usually very large) into a map or list and then do a fuzzy matching and fetch the correct value. The thing is user should be aware of the right value or some sort of filtering, so that the sql query can be generated properly. Also, I would do simple cleaning like lower casing the values.

  • @SabZuso
    @SabZuso 12 วันที่ผ่านมา

    Thank you for the video! An unimportant question: why do you start your prompt with "Please"?!

    • @TwoSetAI
      @TwoSetAI 12 วันที่ผ่านมา

      @@SabZuso haha sure you don’t have to ;)

    • @MehdiAllahyari
      @MehdiAllahyari 12 วันที่ผ่านมา

      LOL... I was trying to be nice to chatGPT ;). You definitely do not need to use words like "please", etc. In fact, there is a research paper that shows if you use words like "i will tip you" or "you're going to be penalized if you dont do it correctly", the LLM tends to be working better. Here's the paper: arxiv.org/pdf/2312.16171v1

    • @RedCloudServices
      @RedCloudServices 11 วันที่ผ่านมา

      Using “Please” can often produce improved results from an LLM - sounds crazy but true

  • @dahiruibrahimdahiru2690
    @dahiruibrahimdahiru2690 12 วันที่ผ่านมา

    How would you implement an agent? Do you recommend going with frameworks that provide agents + tools?

    • @MehdiAllahyari
      @MehdiAllahyari 12 วันที่ผ่านมา

      It's more of a preference! I personally don't like agentic frameworks like autogen, langraph and crewAI. MetaGPT is really good and now that OpenAI has introduced Swarm, which is pretty light and almost a few classes and objects has become my favorite.

  • @dahiruibrahimdahiru2690
    @dahiruibrahimdahiru2690 12 วันที่ผ่านมา

    I did something similar to this, but connected mine to a MongoDB database. Used Gorilla LLM to get the syntactically correct query to run

    • @MehdiAllahyari
      @MehdiAllahyari 12 วันที่ผ่านมา

      Yes, that's also a great setting. :)

  • @happymeatbeer9925
    @happymeatbeer9925 12 วันที่ผ่านมา

    Hi Mehdi and Angelina. You are doing great job,really appreciate your content. Can you cover some content on finetuning llm. Always hear voices like finetuing does not have good outcome... or you may need to fine-tune your own model for specific use case like law..... I am wondering how far can fine-tune go these days😅

    • @TwoSetAI
      @TwoSetAI 12 วันที่ผ่านมา

      @@happymeatbeer9925 hey do you have a specific use case? Let us know!

    • @MehdiAllahyari
      @MehdiAllahyari 12 วันที่ผ่านมา

      Thank you! Fine-tuning nowadays has become very straightforward with several frameworks and the key factor boils down to having a very high quality dataset. Fine tune depends heavily on the use case and if the accuracy of the llm is not good enough (i.e. prompting and RAG may not work well) or the task is very specialized. But in the future if we find a good use case, we'll certainly do a video for it.

    • @happymeatbeer9925
      @happymeatbeer9925 12 วันที่ผ่านมา

      @@TwoSetAI Thank you very much for your reply. I am a software engineer with no background of transformer or model training etc. I have been studying and creating rag based system for quite some time and I kind of know what to expect when I hear a new approach of doing rag. I am expecting, to better enhance the system or create a unique app for specific field,I may need fine-tuning,(do not have use case yet). The coding is not a problem,I tried OpenAI fine-tuning when they just released it (limited data and strange outcome). But unlike Rag I don't know what to expect from fine-tuning a model and I don't find many content about fine-tuning. Maybe next step I will have to search for thesis or do some costly experiment for this. If you have any thing about this,it would be great.

    • @goggledtech47
      @goggledtech47 11 วันที่ผ่านมา

      ​@@MehdiAllahyarihi, thanks for the knowledge you share. I am also interested in knowledge about fine tuning, A great use case would be in the the healthcare or law field, where the documentation of each is done very formally and also a lot of complex terms are used. Building upon a custom fine tuned model to aid in these areas would be great. There's MedLM by google ai but its very private and having an open-source or fine tuned model would be better.

  • @abbathdoom
    @abbathdoom 13 วันที่ผ่านมา

    Excellent explanation thx ! :)

  • @appliedskill
    @appliedskill 13 วันที่ผ่านมา

    Thanks for sharing great explanation on Agentic AI

  • @joyalajohney467
    @joyalajohney467 13 วันที่ผ่านมา

    Super underrated channel. Just binge watched 5 videos today!

  • @Mars.2024
    @Mars.2024 13 วันที่ผ่านมา

    Thank you for clear explanation

  • @eywu
    @eywu 15 วันที่ผ่านมา

    Great interview! Really insightful information

  • @HaseebAshraf-q3q
    @HaseebAshraf-q3q 20 วันที่ผ่านมา

    I have a huge company policy document that I want to create a knowledge graph for, how do I define labels for that? or is it better to do it without? If yes can you please guide me how to go about it without defining labels?

    • @MehdiAllahyari
      @MehdiAllahyari 20 วันที่ผ่านมา

      By labels I assume you're talking about entity names. Those are things that you should already know or have some common sense about. So you can start there or manually create a few and use LLMs or some other model to extract/generate additional ones based on them.

  • @msondkar
    @msondkar 25 วันที่ผ่านมา

    Will this work if I have JSON data instead of text documents? How to work out contextual embedding for JSON chunks?

    • @MehdiAllahyari
      @MehdiAllahyari 25 วันที่ผ่านมา

      It depends on the json data. What is the use case for json? what kind of json data do you have?

    • @msondkar
      @msondkar 21 วันที่ผ่านมา

      @@MehdiAllahyari for example, a nested JSON that contains product details.

    • @MehdiAllahyari
      @MehdiAllahyari 21 วันที่ผ่านมา

      @@msondkar If product details are large enough that you need to chunk them, then you can consider each product as a document and when chunking it, enrich the chunks with extra context from the document. otherwise you can simply have the entire product details as one chunk. I haven't seen your data set, but as long as your documents are large that need to be chunked, then you can use this approach.

  • @TwoSetAI
    @TwoSetAI 26 วันที่ผ่านมา

    Our RAG live course is coming up soon, and as a way of giving back to our amazing community, we're offering you 15% off. Just use this link: maven.com/angelina-yang/mastering-rag-systems-a-hands-on-guide-to-production-ready-ai?promoCode=TwoSetAI We'd love to see you there! 🎉 In the course, you'll have the chance to connect directly with Professor Mehdi (just like I do 😉 in the videos), and you can even ask him your questions 1:1. Bring your real work projects, and during our office hours, we'll help you tackle your day-to-day challenges. This course is for: 01 👇 𝗔𝗜 𝗘𝗻𝗴𝗶𝗻𝗲𝗲𝗿𝘀 & 𝗗𝗲𝘃𝗲𝗹𝗼𝗽𝗲𝗿𝘀: For AI engineers/developers looking to master production-ready RAG systems combining search with AI models. 02 👇 𝗗𝗮𝘁𝗮 𝗦𝗰𝗶𝗲𝗻𝘁𝗶𝘀𝘁𝘀: Ideal for data scientists seeking to expand into AI by learning hands-on RAG techniques for real-world applications. 03 👇 𝗧𝗲𝗰𝗵 𝗟𝗲𝗮𝗱𝘀 & 𝗣𝗿𝗼𝗱𝘂𝗰𝘁 𝗠𝗮𝗻𝗮𝗴𝗲𝗿𝘀: Perfect for tech leads/product managers wanting to guide teams in building and deploying scalable RAG systems

  • @TwoSetAI
    @TwoSetAI 26 วันที่ผ่านมา

    Our RAG live course is coming up soon, and as a way of giving back to our amazing community, we're offering you 15% off. Just use this link: maven.com/angelina-yang/mastering-rag-systems-a-hands-on-guide-to-production-ready-ai?promoCode=TwoSetAI We'd love to see you there! 🎉 In the course, you'll have the chance to connect directly with Professor Mehdi (just like I do 😉 in the videos), and you can even ask him your questions 1:1. Bring your real work projects, and during our office hours, we'll help you tackle your day-to-day challenges. This course is for: 01 👇 𝗔𝗜 𝗘𝗻𝗴𝗶𝗻𝗲𝗲𝗿𝘀 & 𝗗𝗲𝘃𝗲𝗹𝗼𝗽𝗲𝗿𝘀: For AI engineers/developers looking to master production-ready RAG systems combining search with AI models. 02 👇 𝗗𝗮𝘁𝗮 𝗦𝗰𝗶𝗲𝗻𝘁𝗶𝘀𝘁𝘀: Ideal for data scientists seeking to expand into AI by learning hands-on RAG techniques for real-world applications. 03 👇 𝗧𝗲𝗰𝗵 𝗟𝗲𝗮𝗱𝘀 & 𝗣𝗿𝗼𝗱𝘂𝗰𝘁 𝗠𝗮𝗻𝗮𝗴𝗲𝗿𝘀: Perfect for tech leads/product managers wanting to guide teams in building and deploying scalable RAG systems

  • @TwoSetAI
    @TwoSetAI 26 วันที่ผ่านมา

    Our RAG live course is coming up soon, and as a way of giving back to our amazing community, we're offering you 15% off. Just use this link: maven.com/angelina-yang/mastering-rag-systems-a-hands-on-guide-to-production-ready-ai?promoCode=TwoSetAI We'd love to see you there! 🎉 In the course, you'll have the chance to connect directly with Professor Mehdi (just like I do 😉 in the videos), and you can even ask him your questions 1:1. Bring your real work projects, and during our office hours, we'll help you tackle your day-to-day challenges. This course is for: 01 👇 𝗔𝗜 𝗘𝗻𝗴𝗶𝗻𝗲𝗲𝗿𝘀 & 𝗗𝗲𝘃𝗲𝗹𝗼𝗽𝗲𝗿𝘀: For AI engineers/developers looking to master production-ready RAG systems combining search with AI models. 02 👇 𝗗𝗮𝘁𝗮 𝗦𝗰𝗶𝗲𝗻𝘁𝗶𝘀𝘁𝘀: Ideal for data scientists seeking to expand into AI by learning hands-on RAG techniques for real-world applications. 03 👇 𝗧𝗲𝗰𝗵 𝗟𝗲𝗮𝗱𝘀 & 𝗣𝗿𝗼𝗱𝘂𝗰𝘁 𝗠𝗮𝗻𝗮𝗴𝗲𝗿𝘀: Perfect for tech leads/product managers wanting to guide teams in building and deploying scalable RAG systems

  • @TwoSetAI
    @TwoSetAI 26 วันที่ผ่านมา

    Our RAG live course is coming up soon, and as a way of giving back to our amazing community, we're offering you 15% off. Just use this link: maven.com/angelina-yang/mastering-rag-systems-a-hands-on-guide-to-production-ready-ai?promoCode=TwoSetAI We'd love to see you there! 🎉 In the course, you'll have the chance to connect directly with Professor Mehdi (just like I do 😉 in the videos), and you can even ask him your questions 1:1. Bring your real work projects, and during our office hours, we'll help you tackle your day-to-day challenges. This course is for: 01 👇 𝗔𝗜 𝗘𝗻𝗴𝗶𝗻𝗲𝗲𝗿𝘀 & 𝗗𝗲𝘃𝗲𝗹𝗼𝗽𝗲𝗿𝘀: For AI engineers/developers looking to master production-ready RAG systems combining search with AI models. 02 👇 𝗗𝗮𝘁𝗮 𝗦𝗰𝗶𝗲𝗻𝘁𝗶𝘀𝘁𝘀: Ideal for data scientists seeking to expand into AI by learning hands-on RAG techniques for real-world applications. 03 👇 𝗧𝗲𝗰𝗵 𝗟𝗲𝗮𝗱𝘀 & 𝗣𝗿𝗼𝗱𝘂𝗰𝘁 𝗠𝗮𝗻𝗮𝗴𝗲𝗿𝘀: Perfect for tech leads/product managers wanting to guide teams in building and deploying scalable RAG systems

  • @TwoSetAI
    @TwoSetAI 26 วันที่ผ่านมา

    Our RAG live course is coming up soon, and as a way of giving back to our amazing community, we're offering you 15% off. Just use this link: maven.com/angelina-yang/mastering-rag-systems-a-hands-on-guide-to-production-ready-ai?promoCode=TwoSetAI We'd love to see you there! 🎉 In the course, you'll have the chance to connect directly with Professor Mehdi (just like I do 😉 in the videos), and you can even ask him your questions 1:1. Bring your real work projects, and during our office hours, we'll help you tackle your day-to-day challenges. This course is for: 01 👇 𝗔𝗜 𝗘𝗻𝗴𝗶𝗻𝗲𝗲𝗿𝘀 & 𝗗𝗲𝘃𝗲𝗹𝗼𝗽𝗲𝗿𝘀: For AI engineers/developers looking to master production-ready RAG systems combining search with AI models. 02 👇 𝗗𝗮𝘁𝗮 𝗦𝗰𝗶𝗲𝗻𝘁𝗶𝘀𝘁𝘀: Ideal for data scientists seeking to expand into AI by learning hands-on RAG techniques for real-world applications. 03 👇 𝗧𝗲𝗰𝗵 𝗟𝗲𝗮𝗱𝘀 & 𝗣𝗿𝗼𝗱𝘂𝗰𝘁 𝗠𝗮𝗻𝗮𝗴𝗲𝗿𝘀: Perfect for tech leads/product managers wanting to guide teams in building and deploying scalable RAG systems

  • @user-wr4yl7tx3w
    @user-wr4yl7tx3w 26 วันที่ผ่านมา

    why do we need TD-IDF?

    • @MehdiAllahyari
      @MehdiAllahyari 26 วันที่ผ่านมา

      You don't necessarily need tf-idf. It's just a better approach to have two types of search mechanism. 1. semantic search and 2. keyword search. for keyword search tf-idf or BM25 is a natural choice.

  • @dickensyuen6137
    @dickensyuen6137 26 วันที่ผ่านมา

    how can i keep the data base in synchronized with my LOB app say ERP

  • @user-wr4yl7tx3w
    @user-wr4yl7tx3w 26 วันที่ผ่านมา

    excellent format! and great topic.

  • @davidtindell950
    @davidtindell950 27 วันที่ผ่านมา

    What if the length of the 'entire doc' exceeds the 'token-limit' of the Anthropic LLM ?

    • @TwoSetAI
      @TwoSetAI 27 วันที่ผ่านมา

      good question. this is one of the reasons why RAG is relevant and important. Check this post: open.substack.com/pub/mlnotes/p/why-use-rag-in-the-era-of-long-context?r=164sm1&

    • @davidtindell950
      @davidtindell950 27 วันที่ผ่านมา

      Thank U for the relevant Article / Post !

  • @davidtindell950
    @davidtindell950 27 วันที่ผ่านมา

    Yes ! We are obtaining ggood results from contextual retrieval before and now !

  • @PIOT23
    @PIOT23 27 วันที่ผ่านมา

    Great content! It’s nice to see your channel growing too 😁

    • @TwoSetAI
      @TwoSetAI 27 วันที่ผ่านมา

      Thank you for your support! ☺

  • @YuboLi-j8p
    @YuboLi-j8p 28 วันที่ผ่านมา

    Amazing video! A followup question: regarding the judge who decide 0 or 1, what if the judge is incorrect? Any methods about how can we make this role robust enough? Thx!

    • @MehdiAllahyari
      @MehdiAllahyari 28 วันที่ผ่านมา

      Great question! Your judge should be really capable i.e. GPT4o or specialized LLM models for this task. However, even they could potentially make mistakes. Even if judge does miss, what is the worst that can happen? it does an online search and use it to answer the question. So nothing bad will happen. That said, You must evaluate your judge decisions and improve it if necessary!

  • @tink417
    @tink417 29 วันที่ผ่านมา

    is the .ipynb files from this video sourced somewhere for use?

    • @MehdiAllahyari
      @MehdiAllahyari 26 วันที่ผ่านมา

      Here's the link to the code: github.com/mallahyari/twosetai/blob/main/05_sqlite-openai-vanna-vannadb.ipynb

  • @aylinnaebzadeh2650
    @aylinnaebzadeh2650 29 วันที่ผ่านมา

    It could be very useful if you might also provide a code project to show how to build Graph RAGs

  • @MehdiAllahyari
    @MehdiAllahyari หลายเดือนก่อน

    Url for the code is on github. search for "twosetai" on github.

  • @robrever
    @robrever หลายเดือนก่อน

    Awesome video as usual. Can't wait to do the course.

    • @MehdiAllahyari
      @MehdiAllahyari 26 วันที่ผ่านมา

      The course is already out. here's the link: maven.com/angelina-yang/mastering-rag-systems-a-hands-on-guide-to-production-ready-ai

  • @NasserTabook
    @NasserTabook หลายเดือนก่อน

    Excellent tutorial , and in very informative and simple language, can you please share the code with us.

    • @MehdiAllahyari
      @MehdiAllahyari หลายเดือนก่อน

      Thank you! Here's the code: github.com/mallahyari/twosetai/blob/main/13_agentic_rag.ipynb

    • @TwoSetAI
      @TwoSetAI 28 วันที่ผ่านมา

      Its in the video description now!

  • @dailywisdomquotes518
    @dailywisdomquotes518 หลายเดือนก่อน

    can i get GitHub link?

  • @ahmadsuhail2446
    @ahmadsuhail2446 หลายเดือนก่อน

    My only question on this is about the Data Sceurity. Exposing database directly to LLM might be risky. As we have seen many times that certain prompts can some time leak crucial data. So LLM having all the access to the DB without Row Level Security or in this case, any kind of security will be a big big risk to the Organizations

    • @MehdiAllahyari
      @MehdiAllahyari 26 วันที่ผ่านมา

      That's a good point. Of course security is a big deal in every company. There are multiple solutions. One is to have your own LLM, rather than using gpt4, etc.

    • @nglotng
      @nglotng 2 วันที่ผ่านมา

      ​@@MehdiAllahyariThay is not a problem today. And that is the essense of RAG - you can integrate with local, Open source LLM.

  • @manutpunpook9197
    @manutpunpook9197 หลายเดือนก่อน

    beautiful

  • @jennifermiao-b3m
    @jennifermiao-b3m หลายเดือนก่อน

    👍

  • @spicy2800
    @spicy2800 หลายเดือนก่อน

    Thanks for the valuable info

  • @karlfimm
    @karlfimm หลายเดือนก่อน

    Interesting video, but I have issues with the paper. (1) Optimizing each step and assuming that will give the global optimum seems a bit naïve. (2) I'm surprised by the exclusion of chunking strategies like LangChain's recursive chunker. It seems hard to see how a simplistic token count based chunking could ever be better than one that takes into account paragraphs etc (and it's probably faster than sentence level chunking).

  • @vivek-p2t8q
    @vivek-p2t8q หลายเดือนก่อน

    could you please share the code for this

  • @mohammadirmansyah9913
    @mohammadirmansyah9913 หลายเดือนก่อน

    The keypoint in this demo is the pertained model using gpt-3.5 and must be online.