CrewTaskCrewTask V2 - AI-Powered Family OS
CrewTask V2 (current version) - Full-stack AI Family OS. Built solo, part-time, in about 3 months (since Feb 2026) and iterating since, leveraging the foundational work of the first iteration.
Built by Marina Zorina · AI-First Product Engineer & Founding Team Specialist · Based in Playa del Carmen, Mexico
I ship production software end-to-end by orchestrating AI - across any stack. Four projects in React, Python and Next.js, built solo, from a live app to a productized release kit. The output is the proof, not the keystrokes.
What I'm looking for
Looking for a remote founding/early engineer role at an AI-first startup where product thinking and LLM orchestration matter more than boilerplate coding.
Roles I'm built for
Across all four projects the same skill repeats: turn a vague problem into a shipped, reliable system by orchestrating AI. That maps to several roles - each backed by something I actually shipped:
- Founding / early Product Engineer - own a product 0-to-1, from schema and architecture to a live app. Proof: CrewTask, live on Google Play.
- AI Product Engineer / AI Product Lead - design LLM pipelines with evals, structured outputs and failure-mode handling. Proof: CrewTask's multi-tier Gemini orchestration; the research agent.
- Growth / Automation Engineer - build the internal machine: signals to enrichment to automation to AI-personalized output. Proof: the research agent is exactly that pipeline - sourcing, enrichment, scoring, drafting.
- Forward-Deployed / Applied AI Engineer - ship AI that works in the real world and own its reliability. Proof: four projects shipped solo, end-to-end, with monitoring and explicit failure modes.
Remote-only, GMT-5 (Mexico), contractor-ready via Deel - no relocation, no visa.
What CrewTask Is
CrewTask V2 is an AI-powered Family OS for parents who juggle work, kids, logistics, and endless mental load. You speak your chaos as a single voice note, and the system turns it into structured, assignable tasks for the whole family.
This is not just another todo app: it is a multi-tenant, mobile-ready system designed for real families with different devices and asymmetric responsibilities.
My Role
Solo founder · AI orchestrator · System architect · QA gatekeeper
- Designed the product, architecture, and data model from scratch as a solo builder.
- Used LLMs (Cursor, Claude, Gemini) as collaborators, not magic: I orchestrated them, cross-validated outputs, and owned every decision.
- Implemented and verified RLS policies, offline-sync flows, and edge functions with a QA mindset and failure-mode thinking.
- Delivered V2 MVP in 3 months (part-time), while holding a full-time job and raising two kids.
Architecture at a Glance
Frontend & Mobile
- React 19 + TypeScript + Vite + TailwindCSS.
- Zustand for predictable client-side state management.
- Fully transitioned to Capacitor; production builds for Google Play and App Store are in preparation.
- Dual-path voice input: native OS STT (iOS/Android via Capacitor) and Web Speech API (Chrome/Chromium) as primary, with Gemini audio batch fallback for devices without Google services (e.g. Huawei). Leaflet for mapping (cost-optimized at MVP stage).
Backend & Data
- Supabase (PostgreSQL) with Row Level Security for multi-tenant families.
- Deno Edge Functions for AI task parsing, analytics, and business logic.
- Explicit schemas for users, families, tasks, family members, and AI analytics.
- RLS policies designed with a security-first mindset, iteratively hardened using Supabase Advisor insights and real-world QA testing.
AI & Orchestration
- Task parsing: Native STT (OS / Web Speech) feeds voice-to-JSON via three-tier Gemini orchestration scaled by task complexity: flash-lite (fast), flash (standard), pro (complex). Groq llama-3.3-70b as emergency fallback when Gemini is rate-limited or down. Models: (
gemini-2.5-flash-lite, gemini-2.5-flash, gemini-2.5-pro). - Agentic Orchestration & Context Scoping: Development is driven by isolated AI agents with strict rulesets (Frontend, DB, UI/UX).
- Human-in-the-loop QA: Manual cross-validation of AI architectural decisions across different interfaces, leveraging over 10 years of experience in product development teams.
AI Native Tooling & Orchestration
Custom developer workflow built by orchestrating multiple AI agents and remote access tools to enable building, testing, and shipping AI-first products from any device.
MCP Connectors in Practice: MCP (Model Context Protocol) is how I give AI agents direct access to tools — without being the middleman. Claude connects directly to Supabase and Vercel: it runs SQL queries against the live database, reviews deployment logs, and creates tasks in CrewTask autonomously. I also run a scheduled weekly agent that writes tasks to the database without any manual trigger — a real-world example of the agent integration pattern I'm building as a product feature.
Step 1: Phone Connect Setup
Complete step-by-step guide for Windows, macOS, and Linux. Control your Antigravity agent directly from your mobile phone from anywhere in the world.
Step 2: Telegram Bot Setup
This guide helps you install the lightweight Telegram bridge for Antigravity AI. It acts as a headless controller, allowing you to send prompts, switch workspaces, and read artifacts directly from your phone.
Implemented (MVP)
- AI Parser (Multi-tier Gemini 2.5 model orchestration for voice/text-to-task JSON).
- Smart Task Board (Mobile-first Kanban using Tailwind v4 + Zustand).
- Persistent quick-capture - an ongoing notification opens a voice composer; you speak a task and AI parses it, assigns it, and schedules it with priority (Android; no background recording).
- Push Notifications (Multi-channel: FCM for Android/iOS native, Web Push VAPID for browsers, Email via Resend).
- Offline-first sync UI (operability without a network with deferred synchronization).
- Automated Test Suite (Vitest: 327 unit tests in 34 files · Playwright: 88 E2E tests in 20 files · GitHub Actions CI on every PR).
- Error Monitoring (Sentry integration with breadcrumbs and source maps for production debugging).
- Google Play Open Testing - Android app is live in Google Play open testing; full public release is the next step.
6-12 Months Roadmap
- Lock-screen widget - quick task access; under exploration given Capacitor constraints, with the persistent notification shipping as the interim.
- Full Public Release on Google Play - now in open testing (anyone can join); full public release next.
- Hands-free capture via OS assistants (exploring) - Siri Shortcuts / App Intents (iOS) and Google Assistant App Actions (Android), e.g. «Hey Siri/Google, add a CrewTask task». A standalone wake word isn't permitted by either platform, so CrewTask extends Siri/Google instead.
- iOS App - App Store submission and official release (Capacitor iOS already functional with push notifications).
- App Store - official release of the application for iOS.
- Social Component (M-Tribe) - KYC (user verification) and location sharing for verified individuals.
- Two-way Google Calendar sync - temporarily disabled and being re-enabled.
Why This Project Matters
Most "AI productivity" demos stop at a fancy chat. CrewTask is a production-minded implementation:
- A real database with RLS, not a JSON file.
- A real mobile wrapper, not just a desktop web app.
- Edge functions and AI parsing that can be audited and extended.
- Architecture that can evolve into B2C or B2B2C without a rewrite.
I built it to prove that I can take an idea from zero to an AI-powered, mobile-ready product by orchestrating LLMs and still keep the system safe, debuggable, and maintainable.
Decision Stories (What I Can Walk You Through)
- Serverless Migration: I decided to build V2 on a Supabase + Vercel stack to completely offload server maintenance and database scaling to a BaaS. This allowed me, as a solo developer with strict deadlines, to focus exclusively on product business logic.
- Why Capacitor instead of native: One codebase, faster 0-1, enough native capabilities (push, widgets, voice triggers) via plugins without over-investing in Swift/Kotlin.
- How I use LLMs safely: Human-in-the-loop QA, precise engineering prompts, and careful manual verification of the AI-generated architecture.
- When to rewrite: V1 had the core logic but accumulated UI/UX debt — responsiveness issues, design inconsistencies. The real shift was strategic: moving backend operations fully to Supabase BaaS to stop spending cycles on infra and focus entirely on product logic. Everything needed was already available out of the box.
- Supabase Auth Integration: Integration of Supabase Auth with Magic Links and seamless session recovery, drastically improving UX without unnecessary redirects.
How I Build
Across any stack. I orchestrate AI to ship in React, Python and Next.js - architecture, prompt pipelines, cross-model validation and QA are mine; the syntax is the model's job.