Resources to transition to 'Applied AI' role

I have seen quite a few software devs(including me) with little to no background in ML jumping on the genAI train in the last 1-2 years. While this is a good thing and using LLMs with no prior classical ML knowledge helps in some ways it is also a detterent in quite a few ways. Here is how I tried to fill that gap for myself.

Who is this for?

You might be building AI apps or looking to get a job in a team that is dealing with AI stuff, this list of resources will help you do that. The idea is to learn some basic classical ML, understand the common terminology used, get a good understanding of LLMs, etc.

This is more for the devs who have built agents, RAG pipelines, etc but dont have a background in ML. I will also share a few ways to jump start your understanding using reasoning models.

How to jump-start your classical ML knowledge

If you have built a simple RAG app, agentic system, evals, etc there is a lot of things you might have done or used, without knowing what exactly it is. The best way to uncover this, is follow these 3 simple steps:

1. Write down in detail, what did you build, how, what problems you solved, etc. This is your ‘work context‘

   1. Basically write it in as much detail as possible, no fluff.

2. Pick a good ML book which covers a lot of esentials: I used "Designing Machine Learning Systems" by Chip Huyen

3. Copy paste the “Table of Contents” + “work context“ in a model like o3 with a prompt like this

   1. Prompt: “I want to learn what ML concepts i have used and which important ones i have missed. I will provide you with my work context & a list of ML topics. Tell me in detail, which ones have I covered and how to improve them.

      <ML Topics>
      </ML Topics>
      <work context>
      </work context>“

This is a great starting point IMO. You’ll learn a lot of things you have already worked on, giving you great understanding(& confidence :) ).

Pick things you dont understand well, go deeper into it(use the book or chat more with o3) and ask o3 for things that you haven’t covered and get a good understanding of them as well.

Resources

• Designing Machine Learning Systems: Great book to start understanding ML systems and how they are built irl.

• 500+ Real world ML/LLM case studies (pick the ones which are relevant and do a deep dive)

• Great video on how to prep/think for ML System design interview

• 3Blue1Brown Neural Network playlist: Simplest way to get a solid theoretical understanding of Neural Networks + LLMs under the hood. Must watch.

• Neural Networks: Zero to Hero by Karpathy: Great way to get hands on experince with NN.

I think these are more than enough to get started with. Spend good amount of time(I watched 3B1B videos atleast thrice and made detailed notes), understand how they work. Use LLMs to understand, get examples or whatever helps you understand faster!

Pro Tip:

If you are preping for interviews, give the ‘work context‘ to GPT, use it’s voice mode and get it to do mock interviews with you. You can also ask for detailed feedback, how to improve specific points, where you lacked and how to understand it better, etc. This is a game changer IMO.

Conclusion

Transitioning to an Applied AI role doesn't require a PhD in machine learning. By systematically identifying your knowledge gaps and using these targeted resources, you can build the confidence and skills you need. The journey starts with a single ste, so pick a resource and dive in.