I was able to shortlist things I wanted out of my roadmap (see my previous blog). Let’s take them up one by one. Just to recap, these were my points

• What roles are there in AI currently?
• What is the common foundation needed for most of these roles?
• Good resources to study AI topics?
• How to study these topics fast?
• How to keep up with trends in AI?
• Build stuff and showcase and get feedback. How?

What roles are there in AI currently?

It’s good to know what opportunities exist in the market. Sure, learning new things is great. But you don’t want to spend time and effort to study new topics if its not bringing value in some way. This was not much of a rocket science; Grok to the rescue. I was pleasantly surprised to see so many roles. This helps me distill topics and focus on the one that help me in my goal.

What is the common foundation needed for most of these roles?

In general, Probability and Statistics and Linear Algebra are the most important point in this list. I know. Maths is hard, but data in LLMs are nothing but vectors and matrices.

Understanding how LLMs are created and fine tuned using hyper-parameters is good point to follow up.

Deployment, Security, Evaluation, Orchestrator, Costs (Inferencing vs Hosting)

Good resources to study AI topics?

There is no specific list for this one. Pick what works for you. I am a visual learner. I am inclined to YouTubers and podcasts. I like folks that show hands-on along with explaining theory.

• LinkedIn posts by content creators - See
  • Hack - Generally if you open the comments on any of the AI posts, you will see a bunch of OTHER AI content generators. This gives you a good list of people to follow.
• YouTube -
• If you look for it enough, there are a bunch of free ebooks you can find hosted on GitHub, etc. Some authors even publish their books for free, like the ISL book by Gareth James and colleagues. Download at https://www.statlearning.com/. This book is highly recommended for starting out with probability and statistics.

How to study these topics fast?

This is how I do it.

• Use the Pareto principle. 20% only first.
• Start with a top level view. Study the index. Understand and visualise what you’re gonna read in each chapter.
• Think of what use-cases and applications can be applied with each topic.
• Skim through the resource material (PDFs) quickly. Watch YouTube vids on 1.5x speed.
• I create a word cloud on keywords where I’m stuck. I go back to it later.
• Use ChatGPT prompts to study a topic faster. There’s actually a pretty cool way to do this if you look it up. I’ll try and share this technique in another blog. #TODO
• Do the RRR technique. Helps retain things in memory (the human memory here).
• I make notes (in my own words) and mind maps for important topics. I used to write them using a book and pen. But I almost never go back to it since lookup is a mess. Nowadays Google Docs, Miro, LucidCharts work great for me. I’ve been using Notion to create roadmaps and project planning lately.
• Google NotebookLM - I just started using it. Not bad so far. You can shortlist sources you want to study from. The fact that it uses sources directly prevents hallucinations. NotebookLM answers questions or explains topics nicely. I think it uses the Gemini Flash LLM, that makes it super efficient. It has a pretty large context window so you fit in multiple videos and PDFs as sources. They even assemble topics in a podcast like audio. You can intervene and ask questions. It’s like engaging with a live radio-podcast. I thought that was pretty cool.

How to keep up with trends in AI?

It all comes down to how interested you are or wanna be. I’ve been listening to podcasts and YouTubes for a while now. My social medias are now algo-hacked to dumping AI content to me everyday (In hindsight, it does get overwhelming at times).

• I follow various AI content creators (in AI) on LinkedIn. They publish on various and relevant topics from time to time. Most of the content is pretty much what you need, but I send these posts to a WhatsApp/ Slack group I’ve created so I can look it up later.
• I also follow important Tech leaders on LinkedIn. Their statements have the potential to move sentiments in different directions overnight. It gives an idea about where their investments are going, etc.
• Company blogs
• Daily AI News
• YouTube - I’ve been following tutorials. I get to know what is popular and what community is asking for/ working on in AI.
• Podcasts - I listen to these on my way to office and back.
  • YouTube
  • Spotify
• Local meetups in your area. Checkout meetup.com. They usually host on weekends.
• Signup for online webinars. I’ve been getting ads on my insta for webinars organised by industry leaders and startup folks alike. Usually happens on weekends, takes 2-4 hours.
• Form a small group or community with your friends or friends in society who might be interested in similar topics. WhatsApp group works too. Find some like minded folks working in the same domain, preferably different companies so you can share ideas.

Build stuff and showcase and get feedback. How?

The AI landscape is overwhelming. No doubt about it. The trick is not get too bogged down with the theory, but to apply it and build something. It may not be very UX oriented, and that’s fine. The idea is so start making things ASAP, and iterate when needed. Think of this as the MVP-first approach. The fastest way I think to showcase your work right now is -

• Publish on GitHub. Create a detailed README.md.
• Talk about it on LinkedIn, Reddit, etc. Invite collaborators to review.

This should be enough to get you started. You can engage with other users, gather feedback (the feedback loop - for improvements and iterations).

If you have the time, consider making YouTube tutorials or short format reels to showcase more. I personally would stay away when starting out as creating tutorials and reels are a beast of their own as it takes up considerable investment in time and energy.