Work

Product decisions, trade-offs and measurable outcomes.

9 case studies across voice-agent qualification, conversational registration and onboarding automation, API and webhook workflows, advertiser platforms, marketplaces, consumer ranking, experimentation and decision science.

Hypergro

AI Growth & Qualification System

3× demo entries

Reported outcome

AI voice agentsRAG & evaluationConversation design

A RAG-grounded voice-agent journey designed across prompts, conversation flows, microcopy, automation boundaries and human handoff.

Read the full project story

Project summary

The problem

Qualified demand was leaking between interest, product entry and a booked conversation.

Why it mattered

Growth volume did not matter if the product could not identify intent and move the right prospects forward.

My role

Owned product strategy, conversation design, prompts and microcopy, automation boundaries, cross-functional alignment and the release cadence across product, design and engineering.

The hard decision

Ground voice-agent qualification in product and customer context with RAG, while keeping a deliberate handoff to people for high-intent or uncertain conversations.

Impact

The reported result was three times as many demo entries. The work also improved lead-to-demo conversion; a detailed funnel comparison is not included in this public summary.

What I learned

The best AI automation removes friction without hiding the moment when human judgment becomes more valuable.

Hypergro

Performance Creative Automation

50× faster creative production

Reported project outcome.

GenAICreativeCPI / CTR / ROAS

Creative automation that uses CPI, CTR and ROAS to guide what the team makes next.

Project summary

The problem

Creative production moved in long batches while campaign performance changed every day.

Why it mattered

Slow learning creates wasted spend: teams keep producing yesterday’s answer to today’s performance problem.

My role

Led the product system connecting targeting, creative generation, campaign signals and measured outcomes.

The hard decision

Optimize the workflow around feedback quality—not raw content volume—using CPI, CTR and ROAS as signals for the next creative decision.

Impact

Cut creative production time 50×.

What I learned

Generation is a feature. A closed learning loop is the product.

Hypergro

Advertiser Media Operating System

$180K+ managed ad spend

Managed spend: operating scope, not return on spend.

Media buyingCampaign executionMeasurement

One place to manage targeting, creative, budgets and performance across Meta and Google.

Project summary

The problem

Media planning, campaign execution and performance measurement lived in disconnected tools and manual operating routines.

Why it mattered

Fragmentation slowed decisions and made it harder to understand which intervention changed an outcome.

My role

Owned advertiser-side product strategy from media buying through performance measurement.

The hard decision

Give teams opinionated pacing and optimization workflows while preserving control over targeting, creative and budget trade-offs.

Impact

Supported $180K+ of managed ad spend with minimal manual intervention.

What I learned

A good platform turns many operational actions into a smaller number of high-quality decisions.

Hypergro

Three-Sided Ads Marketplace

62% → 95% on-time delivery

Reported outcome

BrandsCreatorsOperations

A shared ads platform for brands, creators and operations, designed to improve revenue and delivery quality together.

Read the full project story

Project summary

The problem

Brands wanted reliable delivery, creators needed a workable experience and operations needed a single source of truth.

Why it mattered

A marketplace fails when one side grows at the expense of the system’s long-term supply quality.

My role

Head of Product for marketplace design, roadmap ownership and cross-functional execution.

The hard decision

Prioritize supply-side quality and on-time delivery alongside revenue rather than treating operations as a downstream service problem.

Impact

Reported on-time delivery increased from 62% to 95%, alongside 48% quarter-over-quarter revenue growth. The two figures describe delivery reliability and commercial growth separately.

What I learned

Marketplace strategy is the design of constraints, incentives and trust—not just matching demand to supply.

Hypergro

Conversational Creator Onboarding

−60% support · −80% time-to-campaign

Reported project outcome.

Registration automationWhatsApp APIsWebhooks

Conversational registration and onboarding through WhatsApp APIs and webhooks, moving creators to a first campaign faster while reducing support.

Project summary

The problem

Creator onboarding depended on manual follow-ups, creating support load and delaying marketplace activation.

Why it mattered

Supply only becomes valuable when it reaches a successful first transaction.

My role

Defined the registration and onboarding journey, conversation design, microcopy, API and webhook automation boundaries, and activation measures.

The hard decision

Automate predictable registration and coordination while keeping escalation paths for exceptions that affected creator trust.

Impact

Reduced support tickets 60% and time-to-first-campaign 80%.

What I learned

Onboarding is not education. It is a designed path to first value.

ShareChat

ShareChat TV Ad Inventory

15M+ daily active users

Product audience: scale context.

VideoAd inventoryMonetization

Ad-slot strategy for a long-form video product, balancing inventory opportunity against viewer experience.

Project summary

The problem

A fast-growing video product needed monetization without degrading content consumption and retention.

Why it mattered

Every new ad slot creates revenue and a potential reason for a viewer to leave.

My role

Defined ad-slot strategy, placement and inventory for ShareChat TV.

The hard decision

Use viewer behavior and product guardrails—not only revenue targets—to decide placement and ad load.

Impact

Defined monetization inventory for a video product scaling to 15M+ DAU.

What I learned

The advertiser outcome and consumer experience are two sides of the same product decision.

ShareChat

Ranking, Discovery & Interest Graphs

100M+ monthly active users in the platform ecosystem

Platform context

Recommender systemsFeed rankingPersonalization

Re-architected ranking and discovery around interest graphs across one of India’s largest social ecosystems.

Read the full project story

Project summary

The problem

Generic ranking wasted attention across a highly diverse 100M+ MAU consumer ecosystem.

Why it mattered

Better discovery compounds into retention, inventory and the strength of the monetization system.

My role

Product manager across feed ranking, discovery, instrumentation and the experimentation portfolio.

The hard decision

Balance exploration and exploitation while measuring retained engagement rather than optimizing a shallow click.

Impact

The published results include improved feed click-through rate and broader event coverage. Platform size describes the operating context; it is separate from the outcomes of a specific intervention.

What I learned

At scale, metric design is product strategy.

ShareChat

Experimentation & Segmentation System

50 tests / year · 19 winners

Experiment portfolio: activity and shipped variants.

A/B testingBigQueryRFM

A structured experiment portfolio plus behavioral segmentation to improve ranking, discovery and monetization decisions.

Project summary

The problem

Product opinions moved faster than causal evidence and event coverage was too incomplete for confident decisions.

Why it mattered

Without reliable instrumentation and a portfolio view, teams optimize isolated metrics and repeat failed ideas.

My role

Ran 50 annual experiments and negotiated roadmap space for an instrumentation overhaul.

The hard decision

Invest in measurement quality before accelerating the volume of experiments.

Impact

Shipped 19 winning variants, lifted feed CTR from 6.1% to 8.3% and event coverage from 54% to 91%.

What I learned

Experiment velocity only matters when the organization gets better at learning.

Mu Sigma

Retail Decision Science

€224M upside · €14M leakage prevented

Revenue opportunity identified, not realised incremental revenue.

RegressionRoot-cause analysisRetail

Analysis of customer and promotion data that led to high-value commercial decisions.

Project summary

The problem

A global retailer needed to understand spend decline across 187K customers and protect holiday economics.

Why it mattered

The visible symptom was revenue movement; the useful product was a defensible decision about what to change.

My role

Decision scientist connecting data, business context, modeling and executive recommendations.

The hard decision

Prioritize interpretable diagnosis over a more complex model that could not clearly guide commercial action.

Impact

Identified €224M in revenue upside and helped prevent €14M in pricing and promotion leakage.

What I learned

Analysis creates value only when it changes a decision.