I'm Deepankar Sharma. Four years on enterprise data products at Cognizant, now doing product work on three live products, including HiMirrorly, an AI reflection product that cites its evidence and abstains instead of guessing. I write the PRD, pressure-test the architecture with working code, and stay accountable for what happens after launch.
An AI consumer product, a research and consulting company site, and a growth agency funnel. Each card is problem, approach, and what happened. Each links to a full write-up.
Self-improvement apps leave you with a private number you cannot show anyone, and AI coaches flatter and hallucinate.
A reflection product where AI does the minimum needed: every score is source-cited and confidence-capped, the engine abstains when it cannot verify, and nothing users write trains a model.
Live at himirrorly.com on a paid plan with a 15-day trial. Early users. Featured in Forbes India, Outlook Business, and The Tribune.
A UAE research and consulting company with ten portfolios, certification training, and a product line needed one site that explains all of it to enterprise buyers without reading like a training brochure.
Made the HI-EI-AI ontology the spine, put the flagship products HiMirrorly and HiCoachly beside the ten portfolios, and led with the LEAD method and each GCC country's own agenda so a buyer sees their context first.
Live at simplilead.com as the company site and product hub. The company reports 28,000-plus learners certified across 12-plus countries; those are company figures, not mine.
An AI-enabled growth agency with twelve service lines needed a site that turns a broad menu into one clear first step for professionals and growing businesses.
Made a free Growth Audit the single entry point, with a human-reviewed report inside one business day, and grouped services under strategy, build, marketing, automation, and analytics with project and retainer models.
Live at catalystsolutionservices.com and taking audit requests. Early stage, so the number that matters is audit-to-engagement conversion, not traffic.
I'm developing Growth OS and inviting interested users to participate in early product research. Register your interest to receive updates and learn about testing opportunities when available.
Name and email are required. Intended use helps me pick research participants; it is optional.
The three questions every US or EU hiring manager asks about an India-based PM, answered up front.
6 to 10 pm IST, which is 8:30 am to 12:30 pm Eastern, 5:30 to 9:30 am Pacific, and 2:30 to 6:30 pm Central European. Plus 7 to 8 am IST for Pacific end-of-day handoffs. Shifts by an hour outside daylight saving.
PRDs, decision notes, and evidence labels are how I communicate. Your morning starts with a written update and a clear ask, not a meeting request.
Contractor through Deel, Remote.com, or direct invoice, or full-time through an India entity. No visa, no sponsorship, no relocation.
I read the SQL, the API contract, and the prompt. I prototype with code to answer feasibility questions before they cost a sprint.
I use AI-led development to move from product thinking to working prototypes faster. The goal is not just to generate code, but to translate user problems, workflows, and product decisions into testable interfaces, flows, and execution-ready systems.
▋Turn product thinking into early UI flows and MVP scaffolds in hours, not weeks.
Use quick builds to pressure-test flow, feasibility, and user value before scaling scope.
Translate PRDs and stories into concrete technical artifacts that surface real implementation questions.
Treat AI-generated code as draft material. Review, test, and disclose what the tool did and what I decided.
Ten connected stages that loop as evidence changes. Each stage answers one product question and has an AI checkpoint that asks whether AI belongs there at all.
I spent four years on the execution side at Cognizant: owning backlogs, clarifying requirements, triaging defects, and moving releases. From there you see exactly where unclear problems turn into rework and where a missing decision costs a sprint.
That is why I work the way I do now: define the problem and the metric first, test the riskiest assumption cheaply, prototype with code when a diagram will not answer the question, and write documents a team can build from without me in the room.
Problem framing, user research synthesis, assumption testing, MVP scoping, and prioritization.
PRDs, user stories, acceptance criteria, NFRs, Definition of Done, and release readiness checklists.
When AI belongs, which task type it is, how it fails, what the fallback is, and how it gets evaluated and monitored.
I reply within one working day. Send the role, the problem, or just a question about the work above.