I find the problem underneath the ask.

Product vision: make Idea Workspace the single place where innovation work can start, be evaluated, progress and reach an explicit outcome. Team members submit ideas; teams run campaigns and hackathons; AI and human evaluation happen in the same system; and promising ideas move through validation, leader review, build, scale or closure.

What I own: product vision, roadmap and prioritization for the platform, working with engineering and stakeholders to shape the workflows behind the broader Idea Pipeline. The product has supported 1,500+ ideas across 50+ campaigns and initiatives.

Starting question: “Can we use ChatGPT to evaluate an idea?” I reframed that into a broader product question: can AI provide a scalable first layer of decision support before scarce leader attention is used?

I built the first prototype, validated it with 100+ senior leaders, took it through senior leadership reviews, and helped turn it into an API integrated into Idea Workspace. It has now processed 1,000+ ideas and acts both as the first evaluation layer and as an additional perspective for human evaluators.

Product problem: two physical experimentation spaces contain multiple zones, devices and eligibility rules, so booking and device access cannot be handled by a simple calendar.

I own the product side of the platform being built to manage space booking, zone eligibility, device provisioning and borrowing, shaping the end-to-end rules and experience with engineering. In parallel, we are identifying reusable patterns that could inform other internal products.

Problem: a powerful forecasting engine already existed, but analysts and planners depended on the team that built it to run simulations, creating turnaround measured in days.

As part of a Product Fellowship alongside a senior product manager, I worked on the user-facing layer that made simulations accessible on demand, then focused on rollout, onboarding, user requirements and the next iteration. Turnaround dropped by up to 90%.

Problem: team members needed a simple way to understand where they stood on AI fluency and what they should learn next. I built a lightweight AI-assisted assessment product that maps responses to personalized learning pathways.

More than 1,200 team members used it. User and leader feedback drove additions such as result exports, Slack sharing and aggregate tracking through Power Automate.

Problem: broadcast email and channel posts were easy to send but often easy to ignore. I built a small UI connected to Slack workflows where a user can paste a recipient list, format it and trigger individualized 1:1 outreach at scale.

In tracked use cases, conversion improved by 300–400%, and the tool was reused by teams across India and the U.S. The point was not the complexity of the software; it was removing friction from a repeated communication job and measuring whether behavior changed.

The original onboarding idea was a week-long live program. Looking at the underlying job changed the solution: a semi-autonomous five-week journey, beginning before joining and continuing after, largely orchestrated through Power Automate.

The pattern I care about is not “build software.” It is solve the job with the least unnecessary machinery.