Emily Yeh is an AI-native product leader who turns what's newly possible with AI into products.

Emily Yeh
10M+
Users reached
12 to 800
Company growth, as founding PM
0 to 1 and growth
AI, fintech, B2B2C
2 iOS apps
On the App Store
Computer Science
B.S., Carnegie Mellon

Selected work

Bolt Founding PM (grew 12 to 800 people) · pre-PM and 0 to 1 · consumer checkout and account experience
Strategy 1-Click checkout, company North Star pivot
Re-bet the roadmap from endlessly A/B-tested guest checkout to a logged-in, 1-click flow.
+10% checkout conversion

Guest checkout had been A/B-tested to a local maximum. Every incremental test came back noise.

The number nobody was acting on: logged-in shoppers converted roughly 50% higher. The ceiling wasn't the form, it was the account. So I made the case to stop optimizing a metric the org already trusted and re-bet the roadmap on identity and a logged-in, 1-click flow.

The hard part wasn't the build, it was the argument. It became a company strategy shift rather than a feature test, and it re-pointed how the team bet the roadmap.

Outcome
+10%
Overall checkout conversion
Insight
~50%
Higher conversion for logged-in 1-click versus guest checkouts

Founding PM. Made the case to leadership, sequenced the roadmap, shipped with the checkout and identity teams.

Growth Bolt ID, a universal identity layer
A shared identity and verification network across merchants that other teams built on.
70% acquisition, 10M+ users

Every merchant rebuilt accounts and verification alone, so shoppers started from zero each time.

Identity is worth more shared than owned: one network beats N logins. I built identity as a primitive with its own activation surface, not a checkout side effect, and won merchant trust one integration at a time.

It became a platform primitive other teams and merchants built on, not a single feature.

Outcome
10M+
Users on the network
Outcome
70%
Account sign-up rate
Bolt ID: a verified shopper identity card linked to a network of merchant, fraud, and verification signals

Owned it 0 to 1: the primitive, the activation surface, and the merchant adoption path.

Consumer 0-1 Shopper post-purchase experience
Added an account re-engagement channel and Bolt's shopper experience after checkout from zero: order tracking, SMS updates, and a global order dashboard.
30% opt-in (3x target), 50% engagement

Added an account re-engagement channel and Bolt's shopper experience after checkout from zero: order tracking, SMS updates, and a global order dashboard.

Outcome
30%
Opt-in, 3x target
Outcome
50%
Engagement
Valon Senior PM · mortgage-servicing SaaS
Platform Workflow and task management, built for an agentic future
A no-code platform for servicing operators, composing triggers, conditional logic, and human, system, and AI-driven steps in one flow, on primitives other teams reused.
Replaced Zendesk · 200+ agents, 15+ teams

Servicing work lived in Zendesk queues and spreadsheets no team could see across.

The unit of work was a composable step, human, system, or AI-driven, not a ticket. So instead of one-off tooling per team, I shipped a no-code platform ops teams configure themselves, and migrated live queues without dropping regulated work mid-flight.

It replaced Zendesk company-wide, then was wound down when the company pivoted to AI compliance. The platform worked; the strategy changed.

200+
Agents moved off Zendesk
15+
Teams, discovery to rollout
ValonOS: System of Record, Workflows, and Task Management feeding one unified platform

Senior PM, platform and 0 to 1. Discovery through company-wide rollout.

Ledger Lead PM · insurance data pipeline and dashboards
AI Human-in-the-loop AI corrections
Automated 100% of reinsurance policy and claims data review flows with human-in-the-loop, AI-suggested corrections.
~80% less manual review time

Automated 100% of reinsurance policy and claims data review flows with human-in-the-loop, AI-suggested corrections.

~80%
Less manual review time

AI Product Lab (Independent)

Designed, built, and launched end-to-end to make the world a happier & more connected place.

Clara Wedding planning, run from your inbox. Vendor email becomes a structured dashboard. Beta

An agentic AI assistant: reads vendor emails and turns them into structured dashboards. Human in the loop, you approve every action.

agentic AI · LLM extraction and structuring · dashboards · human in the loop

See a demo →
Clara dashboard: an AI brief, a smart inbox, vendors, and guest RSVPs for a wedding
Closet Shopper Buy less, wear more. An AI stylist that scores a purchase against your whole closet. On the App Store

AI stylist and shopping advisor: learns your taste, shows any item on your body with image generation, and scores a purchase against your whole closet. The core loop is a buy or skip critique, not another feed.

I built the in-store shopping copilot in 2 hours, then used it on a 4-hour trip. It reviewed each piece in under 5 seconds, so I shopped about twice as fast without decision fatigue.

iOS · multi-modal LLM · image-generation try-on · visual embeddings · recommendations

View on the App Store →
Closet Shopper app: the Dress Me outfit builder, the closet grid, and the outfit calendar
Orphea Letterboxd for live theater. A diary and social layer for the shows you see. On the App Store

Talk through a show and it turns your freeform notes into a structured diary entry. Rate performances, follow people whose taste you trust, and get recommendations for what to book next.

iOS · native · social · diary · speech-to-text (STT) · recommendations

View on the App Store →
Orphea app screens: the feed, a show with its rating and review, and a profile

How I work

AI runs through my whole product process as a second brain. It synthesizes, red-teams, and prototypes. The strategy, direction, & insights on human behavior and emotions stay with me.

Oracle for research

AI as an oracle over my own research: it synthesizes my interviews and pushes back on my hypotheses, with exact quotes I can trace to the source.

Red-team the spec

Before the pod sees a PRD, I have AI attack it, the assumptions, the edge cases, the failure modes, so it arrives pressure-tested and data-backed.

Prototype to decide

I prototype in Claude Code and sit next to design and eng to iterate, building a few versions and testing them against each other before picking the best approach.

Two apps on the App Store →

Earlier: Index (acquired by Stripe) and OkCupid. Founded Bella, 15K+ volunteer sessions.

Recognition

Named an "AI Leader to Follow" by Allie Miller, former Head of AI at Amazon 2019
AI & Ethics panelist, Grace Hopper Celebration 2018
Executive 10X Impact Award Bolt
Behavioral-ML research presented at SPSP and featured in New York Magazine 2016

When AFK

I sing with and sit on the board of Harmonia, a barbershop chorus ranked in the world's top 50. Over the years, my love of music has driven my passion for helping more people gain confidence with their voices, teaching 850+ lessons. Off the clock, I'm usually doing my best to spoil my dog.

Emily and her partner with their dog on the City Hall staircase