DeepLearning.AI retweeted
It's been a week since our short course with @DeepLearningAI and @AndrewYNg went live: Tabular Foundation Models for Data Science Agents.
I taught it alongside @FrankRHutter and Eliott Kalfon. It covers four main topics:
· How tabular foundation models like TabPFN predict on a new dataset without training
· Interpreting results
· How these models are actually built (Frank even walks you through training your own nano TabPFN from scratch, on synthetic data)
· And how to hand that capability to an AI agent so it can recognise a prediction task and call the model as a tool.
In the agent lesson, we give Claude Code the TabPFN MCP server and a skill, so the agent decides when a question is a prediction problem and delegates it rather than hand-rolling a model.
If you've gone through it already, let me know what worked and what didn't, especially the hands-on labs.
If you haven't, it's free. Enroll here: bit.ly/4sQZVAA
DeepLearning.AI retweeted
I'm teaching two hands-on workshops on agent harnesses at AI Dev NYC: measure, then optimize.
@AndrewYNg and the @DeepLearningAI team host it at Chelsea Piers in New York on November 30 and December 1.
A harness is the code around the model: its prompts, its tools, how it manages context and when it decides to stop.
Part 1 on November 30: "Your benchmark score is a harness score." With identical weights, Muse-Glimmer-30B costs $1.73 per solved Terminal-Bench 2.1 task in OpenCode and $8.80 in OpenClaw.
In the room, each table runs two harnesses on a 10-task slice of Terminal-Bench 2.1, and we build a shared leaderboard of pass rate and cost.
Part 2 on December 1: "Don't train the model, evolve the harness." A meta-harness is an agent that reads failed runs and patches the harness around a fixed model.
This loop took GPT-OSS-20B from 4.8% to 14.8% on held-out Terminal-Bench 2.1 for $50 of API credits. In the workshop, you run your own loop and watch the patches land.
Bring a laptop!
Thanks @cgeorgiaw for the intro and @amandamork for the invite!
Sessions: ai-dev.deeplearning.ai/speak…
DeepLearning.AI retweeted
Last day of the 7-day Voice AI Builder Challenge with DeepLearning.AI. The leaderboard is stacked and competition is heating up 🔥
Get your submissions in!
Deadline June 30 at 11.59pm PST
DeepLearning.AI retweeted
🎉 @DeepLearningAI is releasing a new course with CopilotKit
Fullstack Agents & Generative UI ✨
Build with AG-UI, A2UI and MCP Apps. Learn how to connect any agent to your fullstack application with shared state & frontend tool calling.
In this new course, Atai & Andrew Ng walk you through:
1. The Generative UI Spectrum
2. Building a Chat assistant with Generative UI
3. Building with A2UI
4. Using MCP Apps
5. Building fully open Generative UI
6. Building blocks for Fullstack Agent Applications
Save this so that you don't forget!
If you are curious about Generative UI, sign-up for the waitlist:
deeplearning.ai/short-course…
DeepLearning.AI retweeted
Day 1!!
@DeepLearningAI
Spice Founder & CEO @lukekim presenting at #AIDev26!
🗓️Wednesday, April 29th @ 3:30 pm, Stage 3
'Why Every AI Agent Needs Its Own Data Stack'
AI agents break the centralized data and AI model, where a monolithic platform serves disparate workloads. Instead, every agent needs its own sandboxed, secure, and modern data stack. In this 30-minute presentation and demo, we explore why this architectural shift is critical and the key patterns required to give agents reliable, real-time data.
Thanks to @DeepLearningAI for hosting!
Check out the full AI Dev 26 agenda here: hubs.ly/Q04b_j-_0
AI Dev 26 brings the builders together. We'll be among them.
Meet us there 👇 @DeepLearningAI @AndrewYNg
3,000+ developers. Two days.
AI Dev 26 x San Francisco is where the people building AI come together to discuss and shape the future of software engineering.
Join Andrew Ng and leading experts across the field at Pier 48, April 28-29.
Tickets are selling out. Grab yours today: hubs.la/Q04bqBf_0
DeepLearning.AI retweeted
Document processing isn't document understanding!
OCR reads text. It extracts words from PDFs. But it misses layout, structure, and relationships. A table becomes a wall of text. Reading order scrambles. Spatial context disappears.
Documents aren't just text containers. They're visual systems. The position of a field relative to its label matters. Table boundaries matter. Hierarchies matter. Multi-page flow matters.
Traditional OCR is perception, not cognition. It sees characters but doesn't understand what they mean in context.
Agentic Document Extraction (ADE) approaches this differently. Vision-first architecture that preserves structure. Parsing that maintains relationships between elements. Every extraction grounded to its exact location in the source.
Our free course with @DeepLearningAI walks through why traditional OCR breaks on production documents and how vision-first extraction solves it.
6 hands-on labs covering everything from OCR failure modes to building RAG applications with visual grounding.
DeepLearning.AI retweeted
Learn more about our new short course with @DeepLearningAI, aimed at teaching AI developers and engineers how to architect and implement memory systems that give agents persistence, continuity, and the ability to learn over time. 🧠 social.ora.cl/6017B65FER
DeepLearning.AI retweeted
Everyone's building agents.
Nobody's fixing the document layer.
Agents orchestrate workflows perfectly. They break when documents get complex. Tables misread, layouts vary, values can't be traced.
LandingAI is at AI Dev Day by @DeepLearningAI, April 28-29.
David Park, our Senior Director of Applied AI is speaking on building the infrastructure layer that makes document understanding reliable for agentic systems.
If document understanding is breaking your agentic workflows, come by.
Spice AI is returning to AI Dev!
One of the world's premier technical, hands-on conferences for AI and hosted by @DeepLearningAI, this year's installment is taking place in San Francisco on April 28th & 29th.
Spice Founder and CEO @lukekim will be speaking at the event. Details coming soon.
Check out the agenda and register here: hubs.ly/Q043q7YM0
Why does AI reasoning sometimes look odd during training?
Reinforcement Learning only rewards the final answer, so intermediate reasoning can get pretty strange (from repeated phrases to sudden language switches) and still get a gold star. ⭐
Learn about this and more in the AMD x @DeepLearningAI course on Fine-tuning and Reinforcement Learning for LLMs: bit.ly/4qHIwJd
DeepLearning.AI retweeted
Have you ever needed to manually reformat data? If you’ve ever copy-pasted values from a PDF into Excel at 11PM the night before a deadline, then our course with @LandingAI is for you!
Join us for "Document AI: From OCR to Agentic Doc Extraction,” where you’ll learn to parse and extract information from documents with components like tables, charts, and forms.
Join in today! bit.ly/3Lmuqxf
DeepLearning.AI retweeted
Important new course: Agent Skills with Anthropic, built with @AnthropicAI and taught by @eschoppik!
Skills are constructed as folders of instructions that equip agents with on-demand knowledge and workflows. This short course teaches you how to create them following best practices. Because skills follow an open standard format, you can build them once and deploy across any skills-compatible agent, like Claude Code.
What you'll learn:
- Create custom skills for code generation and review, data analysis, and research
- Build complex workflows using Anthropic's pre-built skills (Excel, PowerPoint, skill creation) and custom skills
- Combine skills with MCP and subagents to create agentic systems with specialized knowledge
- Deploy the same skills across Claude.ai, Claude Code, the Claude API, and the Claude Agent SDK
Join and learn to equip agents with the specialized knowledge they need for reliable, repeatable workflows.
deeplearning.ai/short-course…
DeepLearning.AI retweeted
Just finished the "Agent Skills with Anthropic" course by @DeepLearningAI 🎓
Main takeaway: you can teach Claude custom skills for specific tasks, and it works across their API, Claude Code, and Agent SDK. Makes AI agents way more useful for actual work.
Pretty cool stuff. 😎
DeepLearning.AI retweeted
You had to be there. But if you weren’t, this was AI Dev 25 × NYC.
And next up is AI Dev 26 × San Francisco, our largest event yet.
📅 April 28-29
🎟️ Early Bird tickets ending soon! Get yours here: bit.ly/4aiyfNp
DeepLearning.AI retweeted
Traditional OCR is great... until it hits real documents.
Scanned invoices with merged cells. Medical records with handwritten notes. Contracts with multi-column layouts.
Everything breaks.
The text gets extracted. But the structure disappears. Tables flatten. Reading order scrambles. Relationships between elements vanish.
This is the gap between pulling text and understanding documents.
We built a short course with @DeepLearningAI that walks through closing that gap.
You'll see how document processing evolved from traditional OCR to Agentic Doc Extraction:
→ Why OCR fails on complex layouts
→ How layout detection preserves structure
→ What reading order does for multi-column documents
→ How vision language models capture text and images together
→ How ADE parses documents without losing context
6 hands-on labs. Each one builds on the previous.
By the final lab, you're deploying an event-driven pipeline on AWS that automatically processes documents uploaded to S3 and enables semantic search with visual grounding.
The course is free.
Enroll here: bit.ly/4bpV2rG
DeepLearning.AI retweeted
Excited to announce a free course for Gemini CLI 🚀
The course 𝗖𝗼𝗱𝗲 & 𝗖𝗿𝗲𝗮𝘁𝗲 𝘄𝗶𝘁𝗵 𝗮𝗻 𝗢𝗽𝗲𝗻-𝗦𝗼𝘂𝗿𝗰𝗲 𝗔𝗴𝗲𝗻𝘁 in partnership with the amazing folks over at @DeepLearningAI is now live. 👏
In just over an hour you can go from installation to a @geminicli expert.
This course isn’t just for developers; you'll dive into practical use cases for various tasks such as data analysis, content creation, and personalized learning.
Was so fun to chat about Gemini CLI with @AndrewYNg and hear his positive outlook on the product and open-source community.
👇Learn more and enroll below
#GeminiCLI #DeepLearning
DeepLearning.AI retweeted
Master Gemini CLI and transform your daily workflows at no cost! We've teamed up with @DeepLearningAI to launch our new course: Gemini CLI: Code & Create with an Open-Source Agent.
Enroll today → goo.gle/4qqQ3f0
DeepLearning.AI retweeted
New course: Document AI: From OCR to Agentic Doc Extraction, built with @LandingAI, where I'm executive chairman, and taught by David Park and Andrea Kropp.
Much of the world's data is locked in PDFs, JPEGs, and other documents. This short course shows you how to build agentic workflows that process documents accurately: breaking them into parts, examining each piece carefully, and extracting information through multiple iterations.
Traditional Optical Character Recognition (OCR) captures text but loses context from table headers, chart captions, or reading order of columns. After exploring OCR's limitations, you’ll use LandingAI's Agentic Document Extraction (ADE) framework to process documents. ADE treats pages as visually -- as images -- to parse information and extract fields.
Skills you'll gain:
- Build agents to convert unstructured files into structured Markdown/HTML and JSON
- Use ADE to parse complex data like forms, handwriting, or equations
- Map extracted information to named fields using a specified schema, with bounding boxes for grounding and validation
- Deploy RAG applications with event-driven document processing
Come learn about the best tools for processing documents like financial invoices, medical records, or academic papers intelligently:
deeplearning.ai/short-course…



























