So, I’ve decided to pursue an AI-first career moving forward. But TBH I haven’t really figured out what exactly I want to do in AI, and I’m sure a lot of us are feeling the same way. I was initially (and still am) drawn to the idea of Multi-Agentic Systems using LLMs. I started building automation tools and personal assistants using no-code and low-code tools. But I quickly realised that Agentic systems were just the tip of the iceberg. There are mountains of foundational studies that needed to be done if you want to do well in this field.
Building something from scratch is fun when you follow YouTube tutorials and blogs. But you really need to know things inside out when it comes to building and scaling solutions on your own. I found myself struggling here for a while. I had already built a few apps using no-code tools like RelevanceAI, CrewAI, Make.com. I decided to try to make these same apps from scratch using LangChain, LangGraph and Python. Good, I have a starting point now. But, I soon hit another roadblock. Debugging in LangChain is hard. Sure, I was picking code from ChatGPT, Gemini, Deepseek, Qwen (and now Grok), but running into issues was frustrating.
I eventually started looking at generative AI images (I used to love drawing landscapes and characters and illustrations till high school so this was bound to happen sooner or later) using LLMs. Also to mix and match various LLMs and LoRAs needed some deep understanding of LLMs, their fine tuning and hyper-parameters. I explored roles and responsibilities in MLOps and LLMOPs, but didn’t go far either.
With so many things happening in AI every week, it’s not easy to pick a starting point. What I did was to focus on things that mattered to me most. I wanted to build things (passion projects). So I had to take a huge step back to create a roadmap for myself. These were the considerations I needed for my roadmap
- 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?
These were a really good starting point for me. It helped me organise a lot of chaos into actionable to do lists.
I’ll share more on these points in upcoming blogs.
About Rohit Diwakar
Coder. Developer. Blogger. I'm an AI Agentic Developer Consultant, with 15+ years as a Full Stack Engineer and Cloud Architect for companies like Teradata and JPMorgan Chase. I have expertise in building scalable systems with recent focus on agentic AI solutions using Python, LLMs, and cloud platforms. You can find me on LinkedIn.