About Journey Case Studies Principles Skills Contact
Product Manager · AI, Health-tech & Fintech

Ishwar Varshney

I build products people quietly rely on.

Four years turning ambiguous, high-stakes problems into systems that ship, scale, and earn trust, across conversational AI, health records, and rewards at scale.

Let’s connect
Currently PM · Eka Care Based in Bengaluru, India Experience 4 years
01 · About

I care less about shipping features and more about whether the right thing got easier.

I didn’t set out to be a product manager. I studied computer science, then kept landing on the same question on every team: not can we build this, but should we, and for whom. That question still runs the show four years later.

I’ve worked across healthcare, fintech, and SaaS, enough range to learn that good product sense travels, and that the details never do. A rewards feed at CRED and a triage agent at Eka Care have almost nothing in common, except the part that matters: someone on the other side is trying to get something done, often under stress.

I think in systems and ship in increments. I’m drawn to problems where the constraints are real (regulation, trust, latency, cost) because that’s where product judgment actually shows. I measure before I build, cut before I add, and treat a feature’s second-order effects as part of the feature.

Right now I’m most interested in AI products in health, where being useful and being careful turn out to be the same discipline.

02 · Career Journey

Four years, four rooms, one throughline.

Different domains and stages, from a seed-stage health startup to a fintech unicorn, connected by the same instinct for metrics, users, and the messy middle where products actually get made.

Eka CareOct 2025 – Present · Bengaluru
Product Manager · AI Agents & Platform (B2B2C)
Health-tech

Leading 0→1 conversational AI that automates the patient journey where hospitals still run it by hand.

  • Building AI agents for appointment management and patient triage across WhatsApp and web, integrated with hospital HMIS & EMRs across 25+ hospitals.
  • Shipped a self-serve API Validation Platform so partners can verify integrations themselves, turning onboarding into a repeatable GTM motion.
  • Owning prompt engineering and an evaluation framework to keep agent behaviour testable, not hopeful; drove client growth from 0 to 25+ paying clients.
Conversational AIPrompt engineeringEvalsHMIS / EMRGTM
SocialPlusNov 2024 – Jul 2025 · Bangkok / Remote
Product Manager I · Social Engagement & AI Insights
SaaS

Owned social features and analytics for a developer-facing engagement platform.

  • Drove product development of chat, short video, and social features for the SDK and UI-kit across platforms.
  • Built AI-enabled analytics dashboards tracking growth, engagement, and retention for 10+ B2B clients.
  • Worked shoulder-to-shoulder with design, engineering, and customer success to refine features and lift engagement.
Social SDKAnalyticsRetentionB2B
CREDDec 2022 – Mar 2024 · Bengaluru
Product Analytics · Ads, Rewards & Monetisation
Fintech

Personalized a high-traffic rewards surface without breaking the metrics that hold it together.

  • Shipped cohort-level personalization on the rewards feed: +13% cross-sell and +10% cashback claims with no hit to key check metrics.
  • Introduced game-like reward constructs that lifted redemption by 17%.
  • Ran 20+ A/B tests and reporting automation to keep the experiment loop fast and campaign performance sharp.
PersonalizationA/B testingMonetizationCTR / CPM
Eka CareJan 2022 – Dec 2022 · Bengaluru
Associate Product Manager · Health Records & Insights
Health-tech · Consumer

Made it effortless for people to get their scattered medical records into one place.

  • Built records features for a consumer app with 10M+ installs to store, share, and analyze health data.
  • Shipped a Gmail sync flow (Google APIs) that pulls records in three clicks: 0→100K users, +150% records per user.
  • Ran cohort campaigns hitting 13% CTR vs a 2–3% baseline and 4–6X ROAS across SMS, WhatsApp, and paid.
Consumer growthGoogle APIsActivationLifecycle
03 · Featured Work

Three problems, told the way they were actually solved.

Not feature lists. The decision behind each one. Open any card to see the problem, the constraints, and what moved.

A note on honesty: every impact number below is taken directly from my résumé. The framing of constraints, decisions, and lessons is my own retelling of that work. Read it as narrative, not as a line item.

Problem

Care has gaps between steps: booking, reminders, triage, follow-up. Each handoff is a place to lose a patient, and hospitals absorb the load manually.

Context

Eka Care sells into hospitals running legacy HMIS and EMR systems. Any automation has to speak to those systems and reach patients on the channel they already use.

My role

Product lead on the agent (0→1), owning it from problem definition through integration and go-to-market.

Constraints

Patient-facing messaging with clinical stakes: wrong information isn’t an inconvenience, it’s a risk. The integration surface differs per hospital, and trust has to survive every message.

Decisions

Scoped the agent around explicit journeys (appointments, triage) rather than an open-ended chatbot. Invested early in prompt engineering and an evals framework so behaviour could be tested. Built a self-serve validation platform so partners weren’t blocked on us.

Execution

Integrated across 25+ hospitals on WhatsApp and web, with the validation platform compressing partner onboarding into a repeatable motion.

Impact

25+hospitals integrated0→25+paying clientsNewrecurring revenue line

Lesson

In a regulated domain, the eval harness is a product surface. It’s what lets you move quickly without gambling on trust.

Problem

One rewards experience for everyone means the wrong offers for most people, and a quiet ceiling on cross-sell, claims, and revenue.

Context

The rewards section is a high-traffic surface where personalization, ad revenue, and user experience actively pull against each other.

My role

Product analytics across ads, rewards, and monetization, owning the experiments end to end.

Constraints

Lift monetization without denting the core “check” metrics that define the product. Every change had to be provable, not asserted.

Decisions

Personalized the feed per cohort, added game-like constructs to make claiming rewards worth doing, and reallocated impressions to balance CTR, CPM, and revenue.

Execution

Ran 20+ A/B tests with reporting automation so the loop stayed fast and the read on each change stayed clean.

Impact

+13%cross-sell+10%cashback claims+17%reward redemption0damage to key metrics

Lesson

Monetization wins only count if the guardrails hold. The test design ends up mattering more than the idea.

Problem

Eka Care’s app can store, share, and analyze health data, but only for records that make it in. Manual upload is friction most users never get past.

Context

A consumer app with 10M+ installs, where the value only unlocks once a user’s own data is inside.

My role

APM owning the records ingestion feature and its go-to-market.

Constraints

Reach into private inboxes via Google APIs while keeping the flow to a few taps and the permission ask genuinely trustworthy.

Decisions

Let users link Gmail and pull records in three clicks instead of hunting and uploading, then paired launch with cohort-targeted campaigns.

Execution

Built on Google APIs; drove adoption through lifecycle campaigns across SMS, WhatsApp, and paid channels.

Impact

0→100Kusers on the feature+150%records per user13%campaign CTR (vs 2–3%)4–6XROAS

Lesson

Remove the step people never complete, and adoption mostly takes care of itself.

04 · Product Principles

How I make calls when the answer isn’t obvious.

Not rules. Defaults I return to when a decision is genuinely uncertain and something has to ship.

i.

Users before features

Start from the job someone is stuck on, not the thing we’re excited to build.

ii.

Measure before building

An instrumented guess beats a confident one. Decide what would change your mind first.

iii.

Reduce, then add

Most roadmaps are too long. Often the best feature is the one you take out.

iv.

Systems over hacks

A fix that doesn’t generalize is a debt. Build the mechanism, not the patch.

v.

Ship thoughtfully

Small, reversible releases. Trust is lost fast and rebuilt slowly, so protect it on every launch.

vi.

Careful is a feature

In health and money, safety, privacy, and correctness aren’t constraints on the product. They are the product.

05 · Skills

The toolkit, grouped by what it’s for.

Product

0→1 productRoadmappingDiscoveryPrioritizationGo-to-marketStakeholder management

Research & Analytics

User researchSQLMixpanelA/B testingData analyticsExcel

AI & Platform

Conversational agentsPrompt engineeringEvalsAPI integrationAutomation

Growth & Monetization

PersonalizationLifecycle marketingActivationRetentionExperimentation

Craft

FigmaProduct design senseSpec writingSystems thinking

Domains

Health-techFintechSaaSConsumerB2B2C
06 · By the Numbers

A few outcomes worth keeping.

Every figure here is pulled straight from real, shipped work.

10M+
app installs served with records features
100K
users onto Gmail records sync, from zero
25+
hospitals integrated with AI agents
20+
A/B tests run across rewards & monetization
+150%
medical records added per user
+17%
reward redemption at CRED
4–6X
return on ad spend on lifecycle campaigns
0→25
paying B2B clients acquired
07 · Contact

Building something in health, fintech, or AI?

I’m open to product roles where the problems are hard and the stakes are real. Say hello.

Email ishwarvarshney@gmail.com ↗ LinkedIn in/ishwarvarshney ↗ Location Bengaluru, India