Writing
Ten essays. Each answers a specific question a PM recruiter asks - grounded in a product I actually shipped.
Featured essay
FitWardrobe started as a computer-vision demo. It only became a product when I stopped defending my prototype and started listening to the seven friends who told me it solved the wrong problem.
Answers: How do I discover the right problem?
Read the essay
The moment my technical instinct stopped being an asset and started being a tax - and the three questions I now ask before I let myself open a code editor.
Answers: How did my mindset evolve from builder to PM?
The one-page pre-PRD I run before any spec - and why the "must-have" column is the one that decides whether the product ships on time.
Answers: How do I decide what gets built?
The Credex feature I almost shipped without a definition of "worked" - and the two-column doc I now use so I never do that again.
Answers: How do I know a feature is successful?
A short list of things I built, shipped, then deleted from FitWardrobe - and what each removal taught me about the difference between a capability and a product.
Answers: Can I say "no"?
Why FitWardrobe runs its model on the phone even though the cloud version would be more accurate - a real trade-off table, not a manifesto.
Answers: How do I make AI product decisions?
Trust is a product surface, not a model property. Notes from the AI-First CRM on the four affordances that moved reps from "let me double-check" to "let it log."
Answers: How do I reduce user uncertainty in AI features?
A JETIR-published EV-charging prototype and an ESP32 robot taught me more about product management than any framework - because both had physical failure modes that could not be argued with.
Answers: How did my engineering work influence my PM approach?
Five frameworks I reach for regularly, each shown on a real project - with the version I use, not the textbook version.
Answers: Which frameworks do I actually use - and how?
A short, honest audit of what I would build differently on FitWardrobe if I got to start over on a Monday morning - no revisionism, only the mistakes I can actually name.
Answers: Can I reflect on my work and improve?
What I found after building a LangGraph pipeline for Mithivoices - the orchestration patterns that held up under real usage, and the ones that broke before anyone saw them.
Answers: How do you sequence LLM calls without it breaking?
The one-paragraph difference between ML and agentic AI that most definitions get wrong - and why it matters for the products you are scoping right now.
Answers: Is it just a rebrand, or a new paradigm?
Prompt engineering is table stakes. To build autonomous systems, product managers must understand how to bound action spaces and evaluate agentic loops.
Answers: What does a PM actually need to know about Agentic AI?
Search is shifting from blue links to direct answers. Technical SEO now means making your unstructured data legible to a retrieval-augmented generation (RAG) system.
Answers: How do you optimize a site for an Answer Engine?
Browse by topic
