Case studies

Real work, real outcomes.

Four projects from externships, founding, growth, and analytics: what I shipped, what landed, and what I'd do differently next time.

01Externship · BeReal

RealTogether: a 1:1 social engagement feature

Product Management Extern · Sep 2026

Problem

BeReal's core loop is authentic, but engagement between close friends drops off between daily posts. The opportunity: deepen 1:1 connection without compromising the app's anti-performative ethos.

Approach

Synthesized 100+ qualitative user insights, market research, and competitive analysis to surface pain points. Translated ambiguous problems into actionable requirements, prioritized opportunities with RICE, and defined the product strategy and roadmap for RealTogether: a 1:1 social engagement feature.

Outcome

Delivered a full PRD with user flows, user stories, acceptance criteria, and success metrics, plus an interactive prototype to guide development and measure feature impact.

RealTogether solution slide showing the invite flow — The solution: a shared daily BeReal between two friends.

Watch the RealTogether product pitch.

What I'd do differently

RICE forced discipline, but I leaned on reach estimates I couldn't fully validate. Next time I'd pair the scoring with a quick survey to size the opportunity before committing.

Product strategyPRDRICEPrototypingUser research

02Founder · LiveOne

Building an AI lifestyle platform from zero

Founder & AI Product Manager · May 2026–Present

Problem

Students juggle scheduling, meal planning, fitness, budgeting, and goals across a dozen disconnected apps. Nothing ties it together, and generic AI chatbots don't take action on your behalf.

Approach

Designed multi-agent workflows and end-to-end user experiences that let AI assistants autonomously manage each domain within one centralized platform. Integrated AI APIs and translated recurring user needs into product requirements.

Outcome

Ran iterative product testing with 30 student users, using structured feedback to refine workflows, the roadmap, and the interface across multiple cycles.

What I'd do differently

Early on I built for breadth (five domains at once) when depth in one would have taught me more. I'd start with a single killer workflow and earn the right to expand.

0→1AI productsMulti-agent systemsUser testingRoadmapping

03Growth · Spotnxt

Black Friday go-to-market: a path to $10K MRR

Growth Product Intern · 2026–Present

Problem

Spotnxt, an AI ad platform for Meta and Google, had about five weeks to win US customers before Black Friday. Brands won't switch ad tools during peak week, so the real deadline was Nov 20, and the product has no free plan to hook people with.

Approach

Sized the opportunity ($257.8B US online holiday spend) and chose where to play: US first, led by California with 50,000+ Shopify stores. Used shopper data to show Facebook + Instagram reach both Gen Z and Millennials. Split targets into agencies, D2C brands, and enterprise, then designed a 'free insight first' funnel: build a lead list from Meta Ad Library, send a personalized teaser with 3 fixes from public competitor ads, offer a $99 Black Friday ad check, then upgrade.

Outcome

Presented the plan to leadership with a clear target: 31 customers (10 agencies, 15 Growth, 6 Starter) = $10,069 MRR, plus a dated timeline to the Nov 20 cutoff, a post-holiday campaign calendar to fight churn, and a risk plan. Outreach launches Oct 23, with results to come.

Slide showing 31 customers adds up to $10,069 monthly recurring revenue — The math: 31 customers gets Spotnxt to $10K a month.View the full deck →

What I'd do differently

I built the revenue math from list prices, but I haven't tested conversion rates yet. Next time I'd run a small teaser batch first so the 31-customer target rests on real reply and close rates, not assumptions.

Go-to-marketGrowthMarket sizingPricing mathB2B outreach

04Analytics · Northrop Grumman

Automating obsolescence tracking for 100+ components

Program Planning & Scheduling Intern · Summers 2025 & 2026

Problem

Program teams tracked component obsolescence risk manually across supply chain, engineering, and vendor data: slow, error-prone, and always out of date when leadership needed answers.

Approach

Built an automated obsolescence tracking tool using SQL, Tableau dashboards, and Excel, integrating data across 100+ components. Built Deltek Cobra revenue forecasts translating component risk into program-level financial exposure.

Outcome

Reduced manual reporting by 60%, improved real-time risk visibility, and quantified $5M in program-level financial exposure, guiding stakeholder mitigation decisions.

What I'd do differently

The dashboard succeeded because I sat with the planners before writing a line of SQL. The lesson generalizes: internal tools are products too, and their users deserve discovery.

SQLTableauAnalyticsStakeholdersInternal tools