I sit at the intersection where user behavior, business incentives, and model capabilities collide and I make the call on what ships.

Most PMs treat AI as a feature toggle - flip it on, ship it, move to the next sprint. I treat it as infrastructure.

The difference matters. Features get deprecated. Infrastructure compounds. And right now, the biggest gap in AI product development isn't capability, it's the trust layer between what a model can do and what a user will rely on.

That gap is where I operate.

From Engineering to Product

I came to product through engineering - ETL pipelines, data workflows, systems that fail silently at 3 a.m. That foundation gave me something most PMs don't carry: an instinct for how things break before anyone notices, and the technical fluency to have real conversations with the people building them.

When I moved into product, that instinct became my edge. At Filo, I helped launch an instant tutoring platform in the US that scaled to 120K+ users in six months - not by guessing what users wanted, but by running tight experimentation loops that increased activation 35% and grew signups 125% quarter over quarter. At ProductSquads, I built two AI-native tools from the ground up: an API testing system that cut QA effort by 40% and a document intelligence engine that reduced processing time by 70%.

Different domains, same pattern: define the highest-leverage problem, build the system around it, and measure what actually moved.

Inside the Black Box

In 2025, I started shipping production AI powered by large language models. The capabilities were immediate. The failure modes were worse.

Models hallucinate with confidence. Prompts drift without warning. Latency compounds into cost. Evaluation is ambiguous by design. And trust - the thing that determines whether anyone actually uses what you build - is fragile and slow to earn.

I refused to be another PM bolting a model onto a thin UI and calling it innovation. I wanted to understand the mechanics - how models reason, where evaluation breaks down, how to design feedback loops that improve rather than degrade, and what separates a defensible AI product from a disposable one.

That's what brought me to the University of Washington's MS in Information Management, specializing in Product and AI. Not to add a line to my resume - to build the technical intuition required to make real product bets in a domain where most people are still guessing.

After five years of watching product teams lose signal in fragmented feedback - scattered across Slack, support tickets, CRMs, and sales calls - I built NXTai to solve it. The thesis: when development is frictionless and shipping is fast, the bottleneck shifts from can we build it to should we build it. NXT.ai uses AI to unify customer feedback and quantify the revenue and churn impact behind every feature request - turning the noisiest part of product prioritization into the most rigorous. It's not a startup. It's how I think about every product problem: find the decision that's currently vibes-based, and make it measurable.

The Global Lens

I've built products across India, Singapore, and the United States — for consumers, SMBs, and enterprise clients. That range taught me something that doesn't show up on a spec sheet: the same product means entirely different things depending on who's holding it.

Context shapes adoption. Nuance determines trust. And the ability to read a room — culturally, organizationally, technically — is the most underrated skill in product management.

What's Next

I graduate from UW in August 2026 and I'm looking for product teams where the problems are genuinely hard - teams building intelligent systems that amplify human judgment, not replace it.

If you're working on something where trust, evaluation, and system design matter more than hype, let's talk - sahildua@uw.edu.