From QA to product
My first job was QA at LTI, on a large U.S. insurer's policy systems, premium rating included. When the engine that prices policies has a bug, it misprices a whole book of business, not one policy. That's where I learned to ask what happens when something goes wrong before asking what happens when it goes right.
At Deloitte I moved from QA into product, on digital transformation work for U.S. P&C insurance clients. Then I left a stable job for a full-time postgraduate programme at SPJIMR to make the switch for good.
Three companies, one pattern
The industries look different. Each time the product replaced a slow, manual step: checking trade documents against the rules by hand, remembering when bills were due, typing asset details off a photo.
Traydstream, UK B2B SaaS · Senior Manager, Product
Banks used our AI platform to check trade finance documents against trade rules and flag discrepancies and compliance gaps. I owned the AI/ML roadmap and the client accounts behind most of the company's revenue, and launched the subscription modules that became a large part of recurring revenue.
Xiaomi India, Mi Pay · Product Manager, Consumer Payments
A UPI payments app with 10M+ registered users. I redesigned bill pay, moved registration from app-open to the moment someone actually wanted to pay, and used fake-door tests as one of the main inputs for choosing which merchants and partners to integrate. I also ran an early-stage tap-and-pay pilot on wearables.
Actabl, U.S. B2B SaaS · Product Manager
I own the web and mobile products hospitality businesses use for asset maintenance and operations. I set the roadmap with the C-suite and key customers, took Associate Engagement from idea to launch, owned AI Asset Setup end to end, and selected and rolled out Pendo for product analytics.
I work from Canada for a U.S. company, after a decade of delivering for U.S. clients and employers from India. 10 of my 13 years in tech have been for U.S. clients and employers.
What I'm building now
The thread through my AI work runs from classical ML for banks, to vision and document AI in a SaaS product, to an LLM product I'm building on my own time, where I design the grounding, the evals and the judge. The problems I find most interesting are the ones where trusting the output is the hard part.