Escaping AI Proof-of-Concept Purgatory
And Building Robust, Reliable LLM-powered systems
Welcome back to Vanishing Gradients! This newsletter is where I explore developments in data science, ML, and AI while sharing what I’ve been working on. Whether it’s practical tools, technical deep dives, or practical discussions about the space, my goal is to provide you with actionable and thoughtful content.
Here’s what’s coming up and what I’ve been working on recently:
A free lightning class to help software engineers and data scientists break free from generative AI’s "proof-of-concept purgatory."
A live podcast recording with Ravin Kumar from Google Labs about translating AI research into real-world applications.
A new High Signal episodes focused on building experimentation systems and transforming AI expertise into high-impact consulting.
The launch of Outliers, a podcast series featuring Fireside Chats with leading voices in AI and ML.
Let’s dive into the details.
📖 Reading time: 8 minutes
Escaping AI Proof-of-Concept Purgatory: A Free Lightning Class This Week!
Generative AI applications often begin with excitement—a flashy demo that captures attention and interest. But moving from a prototype to a production-ready system? That’s where challenges arise.
🤖 Hallucinations: How do you handle them?
📊 Monitoring: What metrics matter in real-world deployments?
🧩 Integration: Where do you start embedding LLMs into existing workflows?
To address these challenges, I’m hosting a free lightning class with Stefan Krawczyk (Dagworks, ex-StitchFix). This session will walk participants through the essential steps for creating robust and reliable LLM-powered applications.
🗓 Lightning Class Date: Tuesday, November 19, 7 PM EST
📍 Sign up here: Register for the free session
This class is also a preview of our 4-week course, Building LLM Applications for Data Scientists and SWEs. This course may be for you if you’re interested in learning how to:
🛠️ Build and deploy complete AI systems, not just models.
🔄 Iterate effectively by testing, evaluating, and improving applications.
🗣️ Participate in group discussions and Q&As with experienced practitioners.
☁️ Use over $1,000 in cloud and compute credits to experiment hands-on.
💰 Course Early Bird Pricing: $800 (ends November 30, discounts available; email me for a code!).
📍 Learn more and register for the full course.
This Week’s Events: Generative AI, Data Leadership, and Engineering ML
1️⃣ LIVE Podcast Recording: Generative AI’s Rapid Evolution with Ravin Kumar
Join me and Ravin Kumar, Senior Research Data Scientist at Google Labs, for a live podcast recording exploring how tools like Notebook LM bridge research and real-world applications.
💡 What we’ll cover:
📚 Generative AI at Google: How tools like Notebook LM serve both technical and non-technical users.
🌍 Democratizing AI: The shift from training models to API-driven solutions, and the trade-offs between open-source and hosted tools.
🛠 AI Roles & Personas: How roles are evolving—from fine-tuning researchers to API-focused developers.
🏁 Competitive innovation: How companies like Google are shaping generative AI’s future.
🗓 When: Thursday November 23, 10:00AM EST
📍Where: Register here
2️⃣ The Future of Data Leadership: Building Value, Products, and Careers
In this session, Geetu Ambwani, a seasoned data science leader with extensive experience across healthcare, media, and technology, will share her insights on positioning data science as a strategic function, building impactful data products, and navigating career pathways in data leadership.
🗓 When: Thursday, November 21, 2:00 PM EST
📍 Where: Apply here.
3️⃣ Turning ML and AI into Engineering Disciplines
Join me and Alex Filipchik (Head of Infrastructure, City Storage Systems AKA CloudKitchens) for a fireside chat on how ML and AI are evolving into essential engineering disciplines.
🗓 When: Tuesday, November 19, 12:00 NOON EST
📍 Where: Register here
High Signal: Building an Experimentation System with Ramesh Johari
In the latest High Signal episode, I spoke with Ramesh Johari, a Stanford professor and advisor to companies like Airbnb, Uber, and Bumble. Ramesh shared how organizations can move beyond one-off experiments to build systems that foster continuous learning and innovation.
💡 Key points we explored:
Reducing risk aversion: Why faster, smaller tests improve learning outcomes.
Building cumulative knowledge: Making experiments contribute to a growing understanding rather than isolated findings.
Creating a learning flywheel: How experimentation can drive long-term adaptability and success.
Duncan Gilchrist, my co-producer, reflected on his experiences working with Ramesh:
Ramesh exemplifies the unique combination of technical brilliance and business acumen needed to build successful experimentation systems. His ability to navigate between deep technical ideas and broader organizational goals has shaped my own approach.
🎬 Watch the Clip
Here’s a snippet of Ramesh explaining how to design a learning flywheel and what type of experimentation can lead to a self-learning organization:
💻 You can check out the full episode here or on your app of choice.
Turning AI Expertise into a High-Impact Consulting Career with Jason Liu
In a recent episode of the Vanishing Gradients podcast, I spoke with Jason Liu, an AI consultant who has built $50M+ systems and now helps companies deploy advanced retrieval-augmented generation (RAG) applications.
Jason shared his journey from hourly consulting to securing high-value contracts and emphasized the mindset shifts needed to build scalable AI systems that deliver real business impact.
💡 Key takeaways from our conversation include:
🔍 Moving beyond hourly rates: How Jason transitioned from $170/hour consulting to structuring $50,000+ contracts by focusing on value-driven results.
⚙️ Shifting from deterministic to probabilistic AI: Lessons learned from helping teams adapt to the challenges of probabilistic systems.
🛠 Building scalable AI systems: From recommendation engines at Stitch Fix to RAG pipelines, Jason shared insights on creating production-level AI applications.
🤝 The consultant’s playbook: In a live role-playing session, Jason coached me on client engagement, pricing strategies, and avoiding common pitfalls.
🎬 Watch the Clip Check out how Jason moved from hourly consulting to structuring $50,000 contracts:
This episode should be interesting for anyone navigating the intersection of AI expertise and business consulting—whether you’re a freelancer, in-house professional, or team leader. Check it out here or on your app of choice.
Outliers: Fireside Chats in Podcast Form
Outerbounds has officially launched Outliers, a podcast that brings together many of my Fireside Chats over the years with some of the most influential voices in AI, ML, and data science.
🎙️ Outliers features episodes with:
Hilary Parker: On building machine learning systems that deliver value.
Goku Mohandas: Teaching practical approaches to ML.
Chip Huyen: ML + Infrastructure for Humans.
Michelle Carney: Discussing user-centric ML product design.
Jacopo Tagliabue: Sharing insights on building AI systems for smaller-scale organizations.
The podcast reflects years of conversations aimed at helping practitioners and leaders alike navigate the rapidly evolving AI and ML landscape.
💻 Explore the full lineup and listen here: Outliers Podcast Landing Page
On the Road: Teaching and Learning with the AI Community
Over the past few weeks, I’ve connected with the AI and ML community at PyData NYC and the Generative AI Summit in Austin.
💫 At PyData NYC, I taught a workshop on Building Multimodal GenAI Applications. If you’re curious to dive into the content, you can find the GitHub repo here: Your First Multimodal GenAI App.
I also met Mars Lee, who created a wonderful comic on How to Contribute to NumPy. It’s such a great initiative to make open-source contributions more accessible. Check out the comic here.
🛠️ At the Generative AI Summit in Austin, I:
Participated in a panel with Joe Reis, Jepson Taylor, and Juan Sequeda, discussing AI hype vs real-world value.
Taught a two-part workshop with Stefan Krawczyk on LLM-powered applications, focusing on breaking out of POC purgatory.If you’d like to explore the workshop content, you can find the repo here: AI for Software Engineers.
I’ll be announcing more livestreams, events, and podcasts soon, so subscribe to the Vanishing Gradients lu.ma calendar to stay up to date. Also subscribe to our YouTube channel, where we livestream, if that’s your thing!
That’s it for now. Please let me know what you’d like to hear more of, what you’d like to hear less of, and any other ways I can make this newsletter more relevant for you,
Hugo