Penguin Mail – AI-first email client

Problem

Knowledge workers drown in low-value email; most clients treat all messages the same, forcing users to manually triage updates, promotions, and real conversations.

My role

Led product discovery, defined AI-powered filtering concepts, wrote specs, and built the v1 prototype (Next.js, TypeScript, AI APIs).

AI/ML component

Designed an AI filter where users describe their ideal inbox in natural language; system translates this into routing rules and classification prompts. Evaluated LLM providers, latency trade-offs, and fallback rules.

Product decisions

• • Segmented email into Conversations / Notifications / Bulk based on intent instead of folder rules.
• • Scoped MVP to a Gmail-connected web app with a single "training" surface to avoid overwhelming onboarding.

Impact / validation

• • Ran 10 user interviews with heavy email users; 8/10 preferred the "conversation-only inbox" view conceptually.