Blog
Notes from the build.
How I built what I shipped, and what I learned.
- 015 min read
The PM who codes: where product decisions end and technical decisions begin
A reflection on shipping production code as a PM — when writing the code helps, when it blurs accountability, and the line I try to hold at Better and in freelance work.
- 026 min read
How I built my portfolio website with AI: the actual workflow
The real workflow I used with Google Antigravity and Claude to build adityagangwani.me — prompting strategy, iteration loop, what the AI got wrong, and what I had to fix by hand.
- 035 min read
What AI product management actually looks like, day to day
Not the conference-talk version. What I actually do on the product team for Operate, Better's enterprise AI platform for production ops — discovery, guardrails, and keeping trust in the AI output.
- 044 min read
Registration-to-check-in for grassroots sports
Grassroots organizers take UPI and check payment screenshots by hand. RallyVerse keeps that habit and takes the admin off their plate, from registration to event-day check-in.
- 055 min read
Building FeatureSmith: code-review discipline for ML datasets
Code goes through linters and review before it ships. Datasets usually go straight into training. FeatureSmith v0.4.0 gives them the same treatment: 10 reviewers, a 0–100 readiness score, and a CI gate.