I build AI systems that run real things.

I'm one half of a two-person company. The other half is increasingly a team of nine agents: data agents that pull the numbers nightly, execution agents that ship ad creative and store changes, and a founder report that tells me what matters tomorrow.

Orders grew 8x month-over-month in the first spring, repeat rate 2.5x'd in four months, and the system has survived its own production failures. The case study walks through one.

Founding (and only) designer at Thena.ai, an AI-native customer support platform serving Vanta, Braze, Vercel, Mixpanel, ClickHouse, and Launchdarkly.

Led a complete UI overhaul that 20x'd product usage (600 → 12,000 min/week), shipped onboarding that drove 40% activation and 50% paid conversion, and started writing production React/Next.js to collapse QA cycles from weeks to days.

A hunter agent scans Hacker News and the web every morning and scores topics. A scripting engine drafts word-for-word reels and rewrites them through an adversarial drafter/critic/judge loop until they clear the bar. A design system renders carousels as code.

The same system runs DM automation, pulls Instagram insights daily into a Postgres dashboard, and maintains a formula library of what actually reaches. The account started from practically zero; it's past 8,500 followers now — every number date-stamped.

Log a meal by speaking, or snap a photo and let vision estimate the macros. Realtime voice runs on two providers behind one abstraction; nutrition, cycle, wellness, skincare, and fitness live in one PWA.

One codebase runs both my self-hosted instance and a multi-tenant hosted mode. 284 tests, open source at github.com/deepika-builds/life-os-oss.