Context

Insurance endorsement processing was manual and slow, creating bottlenecks for agents handling thousands of cases/month and requiring heavy CSM involvement for routine document access.

The Bet

I defined the end-to-end product vision for Genome — an internal insurance engine — and prioritised API automation and self-serve tooling to remove operational bottlenecks.

What Made It Hard

Every insurer had a different API format, data schema, and submission protocol. There was no standard. The first month was almost entirely spent mapping the variation — and resisting the temptation to build one generic pipeline that would satisfy none of them.

The Key Insight

The real bottleneck wasn't the automation itself — it was the agent's inability to see the status of any case in-flight. Once we built visibility into the pipeline, agents stopped creating duplicate tickets and CSM load dropped on its own.

How I Built It

• Engineered an end-to-end insurer API automation pipeline from endorsement data to carrier-ready submissions.
• Built automated insurer-format generation pipelines, reducing processing time per endorsement by ~40%.
• Shipped Endorsement Copies on the HR Dashboard for self-serve document access by HR teams.
• Independently prototyped and shipped a production fix using Claude + Figma MCP.

What Changed

• Improved endorsement and agent throughput by 3X.
• Reduced per-endorsement processing time by ~40%.
• Cut CSM support requests by 30% with no additional headcount.

Context

A close family member's cancer treatment spanned years of oncology visits, prescriptions, scans, and handwritten notes — scattered across folders, phones, and memory. Every new specialist appointment started from zero. The burden of reconstructing medical history fell entirely on the patient and family, eating into the time that should have been spent on care.

The Bet

I built the product I needed. A family-facing EMR designed not for hospital workflows or institutional reporting — but for a patient and their family at the moment of care. Simple ingestion, structured timelines, and one-tap sharing at the point of consultation.

What Made It Hard

OCR on decade-old handwritten prescriptions is harder than it sounds. Ink fades, handwriting varies wildly by doctor, and there's no standard format. The second challenge was harder: building something a non-technical family member would actually use without any instruction.

The Key Insight

The insight that unlocked the design: this product isn't used at home. It's used in a doctor's waiting room, on a phone, under stress. Every design decision had to be optimised for that specific moment — not for a calm, deliberate user session.

How I Built It

• Organized oncology and preventive records into a consistent digital timeline.
• Used OCR-led ingestion flows to convert handwritten material into editable records.
• Structured the output for quick retrieval and simple sharing during consultations.
• Designed the experience around real family use instead of institutional hospital workflows.

What Changed

• Digitized 300+ handwritten oncology and preventive records.
• Achieved OCR accuracy above 85 percent across messy source material.
• Reduced doctor prep time from 15 minutes to under 2 minutes.
• Improved patient engagement through easy EHR sharing.

Context

Collections teams were operating across disconnected systems and manual processes, limiting visibility, agent productivity, and downstream liquidity outcomes.

The Bet

I prioritized building an internal collections portal with workflow clarity, better operational instrumentation, and a roadmap shaped directly by agent pain points.

What Made It Hard

Collections is a politically sensitive domain — every change to agent workflows touched someone's performance metric, incentive structure, or way of operating. The technical work was straightforward. Getting 1,500 agents and their managers to actually adopt a new system without a mandate from the top was the real product problem.

The Key Insight

Agents didn't resist the tool. They resisted the feeling that it was tracking them. Reframing the dashboard from a monitoring tool to a personal productivity view — showing each agent their own resolution rate and pending queue — changed adoption entirely.

How I Built It

• Owned backlog and roadmap across approximately 300 user stories in JIRA.
• Worked with stakeholders across business, operations, and tech to sequence highest-impact workflows.
• Shaped the portal around agent usage patterns, resolution actions, and operational visibility.
• Used the product as the backbone for ongoing efficiency improvements rather than a one-time internal tool launch.

What Changed

• Delivered Rs 60L in annual OPEX savings.
• Improved liquidity by approximately 15 percent.
• Enabled 1,500 agents with a more unified workflow layer.

Context

Marketplace sellers needed faster access to working capital in their existing context, but any product had to operate within compliance-heavy lending constraints.

The Bet

I led discovery with Flipkart Marketplace and shaped an embedded product flow that balanced distribution advantage, underwriting needs, and regulated user experience design.

What Made It Hard

The Flipkart seller ecosystem has significant trust debt around financial products — sellers had seen misleading offers before. Every UI decision had to be transparent enough to rebuild that trust, while the compliance team required disclosures that the design team thought would kill conversion. Both were right.

The Key Insight

Compliance requirements, when designed well, become trust signals — not friction. The RBI-required disclosures we were worried about became the most clicked elements on the page once we made them readable and honest.

How I Built It

• Ran partner discovery to map seller needs and integration opportunities.
• Defined the user journey for embedded lending access within marketplace contexts.
• Worked through compliance considerations including DPDP and RBI-aligned requirements.
• Coordinated launch across partner, business, product, and engineering teams.

What Changed

• Generated Rs 10 Cr in revenue within 12 months.
• Expanded distribution through a strong marketplace partner channel.
• Delivered a product that balanced growth with compliance rigor.

Context

Meesho's web experience needed to serve a large and growing Tier-3+ audience while improving acquisition quality, engagement, and eventual orders.

The Bet

I focused on web as a growth lever and invested in recommendations, CRM nudges, and platform performance to improve both acquisition and conversion.

What Made It Hard

Tier-3 web users in India have very different device constraints and browsing behaviours from the urban users most product tools are calibrated for. Standard UX heuristics often didn't apply. A/B test results were counter-intuitive — things we expected to help, hurt.

The Key Insight

On low-end devices with slow connections, speed is not a nice-to-have — it is the feature. Doubling our Core Web Vitals scores didn't just help SEO. It was the single biggest conversion lever we found, because a significant portion of our users were abandoning before the page fully loaded.

How I Built It

• Scaled web experiences for underserved user segments across Tier-3+ markets.
• Launched a product recommendations module to increase relevance and session depth.
• Instrumented Mixpanel to identify funnel drop-offs and behavior shifts.
• Partnered with Google to improve Core Web Vitals and search performance.

What Changed

• Added 6 percent new user acquisition on a 5M user base.
• Improved orders by 2.5 percent on a 2M order base.
• Lifted conversion by 5 percent and increased average session time by 10 seconds.
• Drove 10 percent organic traffic growth.