You're going to create an "Insights Extraction AI App" ("Unstructured Text Data" to "Structured Insights") using a modern stack of tools and AI engineering patterns

An end-to-end deployed (live on the web) application that converts unstructured data (podcast, meeting notes, customer feedback, essays, etc.) and outputs structured insights (tweets, product descriptions, summaries, etc.).

The goal is to build & learn a flexible application structure so you can extend to your use case.

Example applications you'll be able to extend this course's code to: Meeting "Next Steps" extractor, Podcast Summarizer, Customer Survey Trends Extractor, Extract class lessons from lectures, Data stream classifier, Custom apps with your own data

AI Engineering Patterns we'll cover:

* Prompting - The basic 3 step pattern for writing development prompts + optimizers, structured outputs

* Observability - To manage costs, tracing and debugging

* Retrieval/Chains - Adding more context to your LLMs and connecting multiple LLM calls together for a unified outcome

* Model Choice - Tribal knowledge about when to pick certain models over another (spoiler: it depends on your use case)

* Evals - The new style of unit tests for your LLM apps

* Mindset - You'll learn how to approach novel problems with AI

We promise three outcomes with this course:

* You’ll launch an AI-powered app designed to teach you the foundations to build your own tools

* You’ll surround yourself with a new group of friends: other builders who are as passionate as you (and hold you accountable)

* You’ll 2x your likelihood to ship by experiencing first-hand the greatest trait a builder can have: a bias for action

You'll get experience building with a modern AI powered development stack:

* Cursor - Write code, but with ChatGPT by your side

* v0 - Build user interfaces with words, not code

* LLMs - Outsource intelligence to Large Language Models APIs like OpenAI/Anthropic

* LangChain/LangSmith - Orchestrate and observe your LLM calls, ensuring your costs don't run out of control

* Railway/Vercel - Deploy your apps

* Supabase - Simple postgres database

* Zapier - Simple ways to quickly integrate your application with other tools

* Anthropic Artifacts - For prototyping mini applications

"The best way to get good at something is usually to just practice actually doing the thing in question. A lot of very capable people outsmart themselves with complex plans that involve working a lot on fake prerequisites." - Sam Altman 10/27/2024

It’s no longer skill level that's holding you back, it’s just your desire to take a step forward.

Along the journey, you’ll have access to mentors, successful builders that have built apps, a network of like-minded people, creative workshops, and 1:1 sessions.

This program is a fast-paced, highly engaging program, but you don’t need to take time off work to participate. We’ve designed the program to be as flexible as possible for those who work full-time or have busy schedules.

With that said, we encourage you to commit to at least 4-5 hours of your spare time each week, for deep work, feedback groups, and live sessions.

In addition to the course material, this cohort includes:

* Code Templates - All materials shown in class will be available for download so you can continue to quickly build after the class

* Slack - A channel for you to ask questions, connect with other builders, and get support on your projects

* Guests - Each week we'll feature an expert from the industry to talk about AI Engineering

This course is not:

* Learn to be a programmer from scratch

* Learn the absolute basics

* Learn how to fine tune LLMs

On the traditional SWE side, we will build our backend with Python and Postgres, frontend with Next.js, and deployed on Railway. Though this course is taught in python, you have the flexibility to bring whatever tool stack you're comfortable with.

Who you are

* You have a intermediate level understanding of Python + Javascript (React is a plus)

* You're not afraid to roll up your sleeves and learn new technology

* You value shipping something quick

* You want to build AI-powered products

* You're ready to break a big problem down into shippable chunks