- 85
- 102 956
AI Stories Podcast - Neil Leiser
เข้าร่วมเมื่อ 9 ม.ค. 2014
Bringing together some of the best data scientists, machine learning engineers, business leaders and researchers that are at the front of the AI revolution. They talk about their career, how they arrive where they are, give advice and share their vision.
TimeGPT, Nixtla & Forecasting with Max Mergenthaler #53
Our guest today is Max Mergenthaler, Co-Founder and CEO of Nixtla: one of the most popular libraries for time series forecasting.
In this conversation, Max first explains how he got into AI and the lessons he learned from building a couple of tech startups. We then dive into Nixtla and forecasting. Max explains how he founded Nixtla and the different libraries available to build stats, ml and deep learning forecasting algorithms.
We also tallk about TimeGPT, Nixtla's closed-source foundation model for time series. We finally discuss the future of the field along with mistakes and best practices when working on forecasting projects.
🔔 If you enjoyed the episode, please leave a 5 star review and subscribe to the AI Stories TH-cam channel.
To learn more about Nixtla: www.nixtla.io/
Open source librairies (StatsForecast, MLForecast, NeuralForecast): www.nixtla.io/open-source
TimeGPT: github.com/Nixtla/nixtla
Follow Max on LinkedIn: www.linkedin.com/in/mergenthaler/
Follow Neil on LinkedIn: www.linkedin.com/in/leiserneil/
---
(00:00) - Intro
(02:00) - How Max got into Data & AI
(03:44) - Combining Philosophy with Analytics
(09:49) - Lessons from building Startups
(14:00) - Founding Nixtla
(16:23) - Time Series Forecasting
(19:25) - StatsForecast, MLForecast, and NeuralForecast
(26:16) - TimeGPT & LLMs for Forecasting
(34:30) - Why people love Nixtla
(42:34) - Future of Forecasting
(45:51) - Mistakes & Best Practices in Forecasting
(52:12) - Max’s role as CEO
(56:09) - Career Advice
In this conversation, Max first explains how he got into AI and the lessons he learned from building a couple of tech startups. We then dive into Nixtla and forecasting. Max explains how he founded Nixtla and the different libraries available to build stats, ml and deep learning forecasting algorithms.
We also tallk about TimeGPT, Nixtla's closed-source foundation model for time series. We finally discuss the future of the field along with mistakes and best practices when working on forecasting projects.
🔔 If you enjoyed the episode, please leave a 5 star review and subscribe to the AI Stories TH-cam channel.
To learn more about Nixtla: www.nixtla.io/
Open source librairies (StatsForecast, MLForecast, NeuralForecast): www.nixtla.io/open-source
TimeGPT: github.com/Nixtla/nixtla
Follow Max on LinkedIn: www.linkedin.com/in/mergenthaler/
Follow Neil on LinkedIn: www.linkedin.com/in/leiserneil/
---
(00:00) - Intro
(02:00) - How Max got into Data & AI
(03:44) - Combining Philosophy with Analytics
(09:49) - Lessons from building Startups
(14:00) - Founding Nixtla
(16:23) - Time Series Forecasting
(19:25) - StatsForecast, MLForecast, and NeuralForecast
(26:16) - TimeGPT & LLMs for Forecasting
(34:30) - Why people love Nixtla
(42:34) - Future of Forecasting
(45:51) - Mistakes & Best Practices in Forecasting
(52:12) - Max’s role as CEO
(56:09) - Career Advice
มุมมอง: 196
วีดีโอ
Understanding Multimodal LLMs in 5 Minutes !
มุมมอง 20416 ชั่วโมงที่ผ่านมา
From my conversation with Sebastian Raschka, Senior Staff Research Engineer at Lightning AI and bestselling book author. Listen to our conversation here: th-cam.com/video/79F32D9aM8U/w-d-xo.html Link to Sebastian's Blog Post: magazine.sebastianraschka.com/p/understanding-multimodal-llms
Build LLMs From Scratch with Sebastian Raschka #52
มุมมอง 3K14 วันที่ผ่านมา
Our guest today is Sebastian Raschka, Senior Staff Research Engineer at Lightning AI and bestselling book author. In our conversation, we first talk about Sebastian's role at Lightning AI and what the platform provides. We also dive into two great open source libraries that they've built to train, finetune, deploy and scale LLMs.: pytorch lightning and litgpt. In the second part of our conversa...
Code Generation & Synthetic Data With Loubna Ben Allal #51
มุมมอง 641หลายเดือนก่อน
Our guest today is Loubna Ben Allal, Machine Learning Engineer at Hugging Face 🤗 . In our conversation, Loubna first explains how she built two impressive code generation models: StarCoder and StarCoder2. We dig into the importance of data when training large models and what can be done on the data side to improve LLMs performance. We then dive into synthetic data generation and discuss the pro...
He Built an AI Football Coach Assistant & Google Maps Algorithm with Petar Veličković #50
มุมมอง 600หลายเดือนก่อน
Our guest today is Petar Veličković, Staff Research Scientist at Google DeepMind and Affiliated Lecturer at University of Cambridge. In our conversation, we first dive into how Petar got into Graph ML and discuss his most cited paper: Graph Attention Networks. We then dig into DeepMind where Petar shares tips and advice on how to get into this competitive company and explains the difference bet...
Fine-Tuning LLMs, Hugging Face & Open Source with Lewis Tunstall #49
มุมมอง 1.4K5 หลายเดือนก่อน
Our guest today is Lewis Tunstall, LLM Engineer and researcher at Hugging Face and book author of "Natural Language Processing with Transformers". In our conversation, we dive into topological machine learning and talk about giotto-tda, a high performance topological ml Python library that Lewis worked on. We then dive into LLMs and Transformers. We discuss the pros and cons of open source vs c...
MLOps Engineering & Coding Best Practices with Maria Vechtomova #48
มุมมอง 1.5K6 หลายเดือนก่อน
MLOps Engineering & Coding Best Practices with Maria Vechtomova #48
OpenAI, AGI, LLMs Eval & Applied ML with Reah Miyara #47
มุมมอง 8316 หลายเดือนก่อน
OpenAI, AGI, LLMs Eval & Applied ML with Reah Miyara #47
Google, Gemini, Cloud & LLMOps with Erwin Huizenga #46
มุมมอง 1.2K7 หลายเดือนก่อน
Google, Gemini, Cloud & LLMOps with Erwin Huizenga #46
Deep Learning for Autonomous Driving with Andras Palffy #45
มุมมอง 4138 หลายเดือนก่อน
Deep Learning for Autonomous Driving with Andras Palffy #45
Launching 7-Figures AI Product Lines with Franziska Kirschner #44
มุมมอง 5938 หลายเดือนก่อน
Launching 7-Figures AI Product Lines with Franziska Kirschner #44
How He Built The Best 7B Params LLM with Maxime Labonne #43
มุมมอง 2K9 หลายเดือนก่อน
How He Built The Best 7B Params LLM with Maxime Labonne #43
From Biostatistician to DevRel at Deci AI with Harpreet Sahota #42
มุมมอง 1899 หลายเดือนก่อน
From Biostatistician to DevRel at Deci AI with Harpreet Sahota #42
Building AI Startups & Raising Funds with Ryan Shannon #41
มุมมอง 41310 หลายเดือนก่อน
Building AI Startups & Raising Funds with Ryan Shannon #41
Interpreting Black Box Models with Christoph Molnar #40
มุมมอง 86511 หลายเดือนก่อน
Interpreting Black Box Models with Christoph Molnar #40
From English Teacher to MLOps Leader with Demetrios Brinkmann #39
มุมมอง 1.5K11 หลายเดือนก่อน
From English Teacher to MLOps Leader with Demetrios Brinkmann #39
Building Over 1000 Models for Uber with Marianne Ducournau #37
มุมมอง 642ปีที่แล้ว
Building Over 1000 Models for Uber with Marianne Ducournau #37
3 Steps To Win a Kaggle Competition According To World Number 1
มุมมอง 5Kปีที่แล้ว
3 Steps To Win a Kaggle Competition According To World Number 1
World Number 1 On Kaggle with Christof Henkel #36
มุมมอง 19Kปีที่แล้ว
World Number 1 On Kaggle with Christof Henkel #36
Making Algorithms More Efficient with Davis Blalock
มุมมอง 107ปีที่แล้ว
Making Algorithms More Efficient with Davis Blalock
The Story Behind Mosaic ML's $1.3 Billion Acquisition with Davis Blalock #35
มุมมอง 979ปีที่แล้ว
The Story Behind Mosaic ML's $1.3 Billion Acquisition with Davis Blalock #35
Kellin Pelrine - How He Crushed A Superhuman Go-Playing AI 14 Games To 1 #34
มุมมอง 886ปีที่แล้ว
Kellin Pelrine - How He Crushed A Superhuman Go-Playing AI 14 Games To 1 #34
Chanuki Seresinhe - Head of Data Science at Zoopla - Generative AI & AI for happiness #33
มุมมอง 440ปีที่แล้ว
Chanuki Seresinhe - Head of Data Science at Zoopla - Generative AI & AI for happiness #33
Rémi Ounadjela - Data Science at TikTok, Google, Amazon & How to get into Big Tech #32
มุมมอง 430ปีที่แล้ว
Rémi Ounadjela - Data Science at TikTok, Google, Amazon & How to get into Big Tech #32
Barr Moses - CEO of Monte Carlo - DataOps & Data Observability #31
มุมมอง 368ปีที่แล้ว
Barr Moses - CEO of Monte Carlo - DataOps & Data Observability #31
Parul Pandey - Kaggle Grandmaster & ML for High Risk Applications #30
มุมมอง 1.3Kปีที่แล้ว
Parul Pandey - Kaggle Grandmaster & ML for High Risk Applications #30
Marijn Markus - Managing Data Scientist at Capgemini #29
มุมมอง 437ปีที่แล้ว
Marijn Markus - Managing Data Scientist at Capgemini #29
Louis Bouchard - Founder of What's AI & Towards AI #28
มุมมอง 152ปีที่แล้ว
Louis Bouchard - Founder of What's AI & Towards AI #28
Miguel Fierro - Lead Data Scientist at Microsoft #27
มุมมอง 1.7Kปีที่แล้ว
Miguel Fierro - Lead Data Scientist at Microsoft #27
You forget to ask him what's the meaning of life at the end 🗿 (Lex Fridman style)
Haha great idea! Should probably ask this to the next guest on this podcast 😅
Normally, I do not comment. But in this case I wanted to leave a comment. I just stumbled upon your podcast and so far I have listened to two episodes. This is really great work and something I have been looking for for a long time. Great mix of personal and technical questions. And in regards to the content, I especially appreciate that, even though the questions go very deep, you do not need to be a full expert to understand what is going on. So thank you for your great work and I am looking forward to listening to many more episodes of your podcast.
Thanks you so much for you nice comment, you've made my day! It really means a lot to me! Very happy to hear that you enjoy the format of the podcast and that the content is accessible. We have some great episodes that will be released in the coming weeks, hope you'll enjoy :)
When talking about ranking problems what do you mean when you say there isn't an equal interval between different ranks? In customer reviews isn't it fair to assume the difference between 1 star and 2 star is the same as the difference between 4star and 5 star?
I think we must not see this literally as 1 and 2 and so on.. As far as I understood, what he meant was basically let's say there is a product for which a user provided a rating of 1, and for this, you can't predict how better the product must be for someone to give it a rating of 2 for it. Like there is distinction between the consequent ratings but there is no measurable distinction between those. He also quoted another example of predicting the damage caused to a house or a car like damaged, severely damaged. These don't have measurable distinctions.
I loved his book "LLM's from scratch". He is awesome guy, posting interesting on topics blogs.
Yes Sebastian is really inspiring! There is so much to learn from him and his career
Many many Thanks Neil for making this podcast!!
thanks, I like listenig to my favourites authors
It's our pleasure! Hope you enjoyed the conversation :)
Your PhD was in a statistical mathematics and you claim it's not related to data science? What are you stupid?
Yeaahhh finally the new season starts and with one of the best guest possible! :D
Petar is indeed an amazing guest! Thanks a lot for the intro Louis :)
thanks for gat pape and this podcast.
My pleasure! Hope you enjoyed the episode :)
He doesn't really have technical expertise as a sales person and this comes through on any talk he appears on lol
Best explanation so far
Thanks a lot :)
Thank you all for all the knowledge and insights
Thanks a lot to you, very happy to hear that you learned from our conversation with Christof
ok
Wonderful explanation!
Jurgen Klopp becoming Number One on Kaggle after retiring from Football manager carrer.
Real tony stark
Very good episode
Thanks a lot Phil :)
merci pour votre effort !!!!!!!!!!!
I am Louis here. Very inspiring. Subscribed
Thanks Louis, very glad that you subscribed and enjoyed the episode
Very interesting talk. At my company we're also trying to develop a model monitoring system for computer vision tasks. Since during deployment we don't have access to ground truth labels, it's challenging to monitor the model in real-time. The idea predicting errors sounds interesting, but I can already see some issues with it. For example, your poor predictions often come from edge cases. These are rare events so if the original model struggles with them, I can imagine the error predictor would struggle to. Something worth exploring though!
Nice 👺👍🏿
Thanks for bringing Lewis onto your show Neil - great insight. Subscribed!
Very happy to hear that you enjoyed the episode! Definitely a lot to learn from Lewis. Thanks for subscribing :)
Very interesting talk. Thank you for the content.
Thanks a lot :) Happy to hear that you’ve enjoyed the episode
This channel is a gem! Thank you for this content
Thanks a lot Ahmed! Very happy to hear that you’re enjoying AI stories :)
Thanks for the podcast. What was the tool you talked about (Nene ML ??) which uses model trained on error?
The company is called NannyML: github.com/NannyML/nannyml
Amazing to see Reah from OpenAI here!✨
Thanks Jonathan, glad to hear that you’ve enjoyed the episode :)
@joerogan needs him on the podcast let’s get em on there
Yes agree! Mike was an amazing and super inspiring guest
Thank you for you efforts
She went to oxford,bound to succeed in life no matter what.
Incredible episode! So badass👑🎸
Thanks :)
Great Podcast , could you please mention the blog name mentioned during the learning resources part . Thank you
Thanks a lot :) You can find Maxime's course here: github.com/mlabonne/llm-course and Lil'Log is also mentioned: And Lil'Log is also mentioned: lilianweng.github.io/. Is this what you were looking for?
Very nice podcast felt like a detailed summarizer for all the things happening in the llm world
Thanks a lot! Very happy to hear you enjoyed our conversation :) Hope this will be useful to you
Epic conversation!
Thanks a lot Jo!
Tremendous episode. Thanks for sharing knowledge.
It’s our pleasure Luciano! Thanks for your kind comment :)
Great episode
Thanks a lot!! Very glad to hear that you enjoyed our conversation
Amazing episode!
Thanks a lot Phil :)
Amazing short, would love to see more of these
Thanks Nathan! There will be more to come 👌🚀
how to reduce noise in data?
What humble person for all knowledge he has
Yes Christoph is super inspiring! Definitely a lot to learn from him :)
That opening quote is an oxymoron hahaha it’s a hamster wheel! I bet he’s really good at: Legend of Zelda 😊
Twelve labs is a very promising start-up. I hope people around the world will get to know its product and value.
Yes definitely agree! They're building a great product
What brilliant men both of you are! I'm truly grateful for the podcast :)
Thanks so much! Very happy to hear that you’ve enjoyed the conversation :)
Honest to God, this is the first helpful explanation for the validation set I’ve heard; replicate the disparity you expect to see in the test set when extracting the validation set…it’s crazy how simple and sensical that is
Thanks a lot! Glad to hear that you've enjoyed the episode :)
thanks for this
Glad to hear that you enjoyed the conversation :)
Cool episode!
Thanks :)
Very interesting 😊
Thanks a lot ! :)
Am also doing civil engineering, currently in my final year and too deep into AI/ML 🤗
That's amazing! Good luck with your AI/ML journey then 😊
Thanks for sharing ❤
Our pleasure! Glad to hear that your enjoyed the episode :)
Super Insightful!
Thanks a lot, very glad to hear that you enjoyed the conversation :)
Good story but there is one point. He seemed to transition in 2019. Market has become so bad even since last year. I dont think similar transition is even possible now. Even if you have a mathematical background no ML companies hiring Juniors..
*transition to ML
It’s true that the market is getting more and more competitive. But if you really want it, you can do it. Keep applying and keep progressing at the same time. Either by taking an analytics role and then transitioning into Data Science / Machine learning later on or by working on a side project or taking online courses. If you keep progressing and really want it, then you can do it!
@@aistoriespodcast The reality is more harsh than our interests. Define me wanting something, if one devotes his 5-10 years he will even complete another related bachelor and masters along with his own projects to qualify. Thats not justice when someone 2 3 years ago just enters the field with only a few data sciencew projects right? So what you say is not completely honest. Today its not even worth transitioning to this field if you are already on another field but it was 2 years ago... I hate it when youtubers and fake influencers try to market their transition made when its much easier than todays world. They dont know and dont care how harder it is than the past...
@@aistoriespodcast As I said on previous reply just be honest, if youre not really trying transitioning today, you dont know nothing on that.. The internet is full of lying marketing on data science and youre just fooling many people. Just give me statistics of what percent of people trying nowadays got a real job on this field?? (not data analytics jobs or internships that they work you on free and drop you off later..) I dont say its not possible for a fresher if you have a CS degree or maybe statistics BS.. People trying transtion (NOWADAYS!) are doomed I can honestly sayt that unlike you...
@@deniz.7200it is definitely not easy but if you really want it you can do it. Transitioning in the field can mean a lot of things and there are lots of ways to do it. For example, you can start working in a non-technical position at a tech company, learn on the side and transition slowly to a Data Science position.