My journey
No ML degree.
That's the point.
Senior Software Engineer at DoorDash, 12+ years in. I got good at applied AI by building, breaking, and reading real systems. This page is the honest version of how.
The foundation
12+ years of real engineering
System design. Distributed systems. Production incidents at 2am. That plain, unglamorous decade turned out to be the perfect training for building with AI.
The spark
One caching insight, millions of tokens
I was working with Claude at scale. I dug into how prompt caching really works. I found a fix that saved millions of tokens. I presented it to a room of senior engineers who saw the same thing I did: AI engineering is just engineering.
Going deeper
Learning by opening things up
I did not learn vector databases from a course. I opened the claude-mem repo. I used it, broke it, and read its internals until the ideas stuck. That is the method: real systems over tutorials.
Now
Agent Diaries: learning in public
I am learning AI agents the same way. This time I write everything down. The experiments, the failures, the wins. If you are an engineer who feels behind on AI, I am writing for you.
What I believe
- →You don't need an ML degree. You need engineering basics, and you already have them.
- →Most AI content is hype. The answer is proof, not promises.
- →The messy middle is the useful part: the failures, the dead ends, the seventh retry that worked.
- →You can become an AI builder without quitting your job.
I'm doing this next to a demanding job. So can you.