AI Stories

How do you build production grade AI Agents? How do you monitor your system of LLMs with Pydantic Logfire? What is PydanticAI? Why is it different than other agents frameworks? How do we define Agents? What do they do well and what do they still struggle with? Pydantic is downloaded millions of times each day, I had the chance to interview their Founder Samuel Colvin on the AI Stories podcast ?? ?? We talk about Pydantic, Logfire and dive deep into agents with PydanticAI: a Python framework designed to build production grade agents. Want to get answers to the questions above? Sharing links to our conversation in the comments! And you, what framework do you use to build AI Agents? Let me know in the comments!

Had a great time on the AI Stories Podcast with Neil Leiser. We covered my journey into AI, early freelance projects, building NannyML, model monitoring, and the future of NannyML. Neil was an amazing host. He was thoughtful and asked all the right questions to make this a meaningful conversation. Grateful for the opportunity to share these insights. Go and listen to our conversation on your favourite platforms (links in the comments)

11 years at Google working in 5+ teams and 3 different countries! From business analyst intern to senior software engineer, Max Buckley has seen it all! 1?? Started as a Business analyst intern in Marketing Finance in the London office 2?? Moved as a business Analyst in Dublin doing analytics and reporting for support teams in sales. 3?? Said no to a manager that asked him to join his trust and safety team as an Analyst. Ended up saying yes when this manager came with a multi year vision plan for Max to become a software engineer. 3?? Progressed and became a trust and safety engineer 4?? Then moved to Zurich and worked on data engineering projects in the Shopping Eng Prod team. Thought he would stay one year but ended up staying 4 years and half. 5?? Switched to an ML role at Google Cloud AI building models for financial services including an anti money laundering algorithm. 6?? Max is now working in the Server Platform AI research team working on LLM for knowledge management. Want to learn more about Max's career? ?? You can find more info of our entire conversation on the AI Stories podcast in the comments. And you, with how many different teams have you worked in your current company? In how many different countries? Share your story in the comments! #google #machinelearning #llm?

I recently had the privilege of appearing on one of my favorite AI podcasts AI Stories with my friend (and host) Neil Leiser. YouTube: https://lnkd.in/eY4BwD3m Spotify: https://lnkd.in/eZpqv5KK More episodes and info here: https://lnkd.in/eG9scgf5 Check it out!

From business analyst intern to Senior Engineer at Google! Max Buckley has been at Google for 11+ years and what a great career he’s had! Max probably knows Google better than anyone else, across the past 11 years he worked: ?? Across multiple teams including as a business analyst, data engineer, software engineer and ML engineer ?? Across 3 countries: UK, Dublin and Zurich ?? On a wide variety of projects ranging from reporting and ads to building advanced anti money laundering algorithms and RAG systems! New episode of the AI Stories podcast is live ?? Our guest is Max Buckley, Senior Software Engineer at Google! If you want to learn more about Max’s fascinating career and how hard he worked to land his dream position: working with AI, ML and LLMs at Google, then this episode is for you! Links in the comments! How did you land your dream job? Keen to hear your story in the comments ??

Your job isn’t done once you’ve deployed a model in production! This is probably when the most important phase of the ML life cycle comes in: model monitoring! Model performance can degrade over time for three key reasons: 1?? Covariate shift: The data that you see in production is different from the data used during training. This can happened if you train a model on images of dogs taken during the summer but the images that your model is currently seeing are taken in the winter. 2?? Concept drift: The relationship between the input variables and the target changes over time. This is what happened during covid for example: lending money to a restaurant pre-covid could have been safe but lending money to a restaurant during covid was suddenly way riskier. 3?? Data quality issues: Garbage in, garbage out! Very often model performance degrades because the data is receives as input isn’t of high quality (a feature is missing, the null rate suddenly increases, …) So what should you monitor to avoid model performance degradation? I’ll answer this question in another post this week. I’ve also covered this quite deeply on the AI Stories podcast in my conversation with Wojtek Kuberski, CTO of NannyML: The data science monitoring platform! Link to our full conversation in the comments! According to you, what are the key elements that should be monitored after deploying a model in production? Let me know in the comments!?

The are two ways to quickly progress in your career: 1?? Find a great mentor / manager who teaches you a lot? 2?? Do things that are way too hard for you and learn on the spot Wojtek?chose the 2nd option! New episode of the AI Stories podcast is live! Our guest is?Wojtek Kuberski, Co-Founder and CTO at NannyML: THE data science post deployment platform! We talk about:? ?? Model monitoring and how to identify the root causes of model failures ?? Covariate shift vs concept drift ?? How to estimate the performance of a model without access to labels ?? How?Wojtek?founded a freelance company with little experience in the field? ?? Why retraining models does not solve all your issues? And much more What are you waiting for? Go and listen to our conversation on your favorite platforms (links in the comments) And you, what are the key elements that you monitor once you’ve deployed a model in production? Let me know in the comments!

Struggling to Keep Up with AI? We’ve Got You Covered! We added a final bonus lesson dedicated to staying up-to-date in our course “From Beginner to Advanced LLM Developer”. This list is an update of our very own resource list (the one we use at Towards AI) to help you filter through the noise and focus on what really matters. We know that mastering AI isn’t just about coding and algorithms—it’s about staying on top of the updates in the AI landscape. And, most importantly, know which one matters. Why? Staying current is the secret sauce to: ? Gaining a competitive edge by discovering breakthrough tools before they become mainstream. ? Making informed decisions that drive innovation and efficiency in your projects. ? Boosting productivity by leveraging the latest, most relevant insights without drowning in irrelevant updates or wasting time on social media. Here’s what’s inside our updated resource list (don't follow them all, just pick your favourite style): ? 10 Newsletters – Curated to deliver high-impact AI insights right to your inbox without overwhelming you including friends like Shawn swyx W, Lior Sinclair and others). ? 19 Influential AI Leaders & Thought Leaders – Follow the voices that are shaping the future of AI. ? Diverse Communities & Forums – Engage in and learn from discussions across platforms like Discord, Reddit, LinkedIn, X, and Facebook. ? 9 Podcasts – Real-world AI applications and trends with in-depth, engaging discussions (including friends hosting Machine Learning Street Talk (MLST) (Tim Scarfe) Latent Space Podcast, Neil Leiser (AI Stories), Lex Fridman...). ? 10 YouTube Channels – Visual content that breaks down complex AI topics into digestible, actionable insights (including Yannic Kilcher, Letitia Parcalabescu, Shaw Talebi...). Check it out here: https://lnkd.in/dDXF2tGZ What are your go-to AI resources? Let us know if we should add any to the guide! #AI #LLM?

- "Do you think we will reach AGI?" - "Yes, I am highly confident!“ I released last week my conversation with Thomas Scialom, PhD, Senior Staff Research Scientist at Meta on the AI Stories podcast. We dive into GenAI and talk about the Llama models, AI agents, RLHF, AGI, ASI and much more. Here is a snapshot of our conversation. Want to learn more? Sharing some links to the full episode in the comments :) And you, do you think that we will soon reach AGI? When?

What are agents? How do we build them efficiently? How close are we to AGI? What are the key differences between Llama 1, Llama 2 and Llama 3? Very very excited to release this new episode of the AI Stories Podcast! Our guest this week is Thomas Scialom, PhD, Senior Staff Research Scientist at Meta. Thomas lead the development of Llama 2, the postraining development of Llama 3 and contributed to other models and papers like Galactica, ToolFormer, Code Llama and much more! If you want to learn more about LLMs, Agents and the future of GenAI, then this episode is for you! We dive into:? ?? How Thomas lead the development of Llama 2? ?? AI Agents and the future of the field ?? Artificial general intelligence (AGI) and Artificial super intelligence (ASI)? ?? RLHF ?? Llama 1 vs Llama 2 vs Llama 3 ?? How Thomas managed to publish around 20 academic papers during his PhD And much more! What are you waiting for? Go and listen to this episode on your favourite platforms ??: Youtube: https://lnkd.in/dFQag7AJ Spotify: https://lnkd.in/d_vRDmbT Apple Podcast: https://lnkd.in/dkDqByVa I plan to record more episodes around AI agents in 2025. Is this a topic that you want me to dive more into? Let me know in the comments!