Nobody hands you a title that says "AI Engineer" and a clean starting line.
For most of my 15 years, I was a full-stack and fintech engineer — React on the front, Node.js or python or php or sometimes little Java on the back, wealth management and banking systems in the middle. AI wasn't a job description yet.
It was a feature you bolted on if the business case justified it.
That changed gradually, then all at once. I didn't pivot into AI. AI showed up inside the systems I was already building, and the skill that mattered most wasn't a new framework — it was being able to wire a model into something people would actually use, explain, and trust.
Here's what the work itself taught me, written down so it doesn't disappear into another closed tab.
The model's output is not the product
One of the first things I built was a banking proof-of-concept: a simple linear regression predicted a user's spending for next month, and an LLM turned that number into a sentence a person could act on. Nothing about the math was exotic.
- The lesson was that a number on a dashboard doesn't change behavior — a sentence does.
- The regression did the calculating.
- The model that actually mattered was the one translating a prediction into something a human could use.
Explainability isn't optional, it's the whole job
The project that taught me the most was a fraud detector built on XGBoost with SHAP explainability.
Banks don't deploy black boxes, and for good reason — a "92% fraud probability" with no justification is not a usable answer, it's a liability.
Once I added SHAP attributions and had an LLM translate the feature weights into plain language, the project stopped being a notebook exercise and started looking like something a compliance team could actually sign off on.
That gap — between a model that scores something and a system that can justify the score — is where most of the real engineering lives.
Fixing the same mistake five times is a signal, not a nuisance
Anyone working with AI coding tools knows the pattern: you correct the same mistake five times before it sticks.
Eventually I stopped correcting it live and started writing the correction down once, as a reusable skill file the assistant would apply automatically going forward.
That turned into two npm packages — skillforge-ai and reactforge-ai-skills — built specifically to teach AI assistants standards instead of re-explaining them every session.
It's a small idea, but it scales in a way that live corrections never do.
The fundamentals didn't change, the substrate did
Feature engineering became prompt and context engineering. Model validation became evals.
Explainability was never optional in finance, and it isn't optional for LLM-driven decisions either.
None of the underlying engineering discipline is new — what's new is how fast the substrate moves under it, which means the engineers who keep their fundamentals tight are the ones who can actually keep up
The part that doesn't show up in a changelog
I've built most of this while navigating a dependent visa — learning exactly what work I could and couldn't take on — and rebuilding momentum after an injury that could have ended the whole thing. None of that is in a GitHub repo.
But it's the reason persistence, more than any specific framework, is the actual skill underneath everything above.
The technical work is the visible part.The discipline to keep showing up for it is the part nobody puts in a release note. If there's one habit worth repeating: don't wait until something is polished to write about it.
Write down what you built this month, even the broken parts.
The polish comes later, if it comes at all — the record is what compounds.
Our Mission
Helping people learn, build, grow, and succeed through technology, education, and community.
“Knowledge becomes more valuable when it is shared.”
“Learn. Build. Share. Grow.”
❤️ Your support, likes, shares, and feedback motivate us to keep creating free educational content for everyone.
Thank you for being part of the journey! 🌱
Built with love by Kirti Kaushal — 15 years of engineering, finally building in public.
Kind is Cool 🙏 share your knowledge
Thank You for Reading ❤️
Thank you for spending your time with us and being part of our learning community.
🌟 Kind is Cool 📚 Share Your Knowledge
🚀 Help Others Grow
If you enjoyed this blog, tutorial, or Instagram post:
👍 Give it a Like
💬 Leave a Comment
🔄 Share it with someone who might find it helpful
⭐ Follow us for more content on AI, Coding, GitHub, Career Growth, and Continuous Learning
💻 GitHub: https://github.com/kirti
Top comments (0)