RELAY

Small-team task extraction AI — turns messy chat into structured work, automatically.

CategoryAI PRODUCT · MOBILE + API

Year2026

StatusIN ACTIVE DEVELOPMENT

RoleFounder · Full-Stack & AI Engineer

The Problem

Every small team (2–10 people) has been there. Someone writes in the group chat: "Let's launch Friday. Rahul handles backend, Sarah does the landing page, and I need to review analytics." Then Friday comes, and half of it didn't happen because nobody turned that chat into actual tasks. Important work gets buried in Slack/Teams/Discord messages faster than you can say "can you send me that again?"

The Idea & Solution

Relay sits where your team already talks, quietly listens for action items, deadlines, and ownership — then turns those throwaway sentences into a real task board. Capture text → AI understands → team reviews → tasks created with assignees and due dates → team executes. No more manually turning conversations into Jira tickets.

How I Built It & Architecture

A React Native cross-platform mobile app (iOS + Android) pairs with a FastAPI backend that orchestrates Groq LLM for ultra-fast structured task extraction. Authentication, workspaces, projects, team members, tasks, assignments, due dates, status tracking, and a project dashboard — all with a PostgreSQL relational core.

SYSTEM TOPOLOGY

[ React Native Mobile App ]
     │  (JavaScript · iOS/Android)
     ▼
[ FastAPI Backend Service ]
     │
     ├──────────────────────────────────┐
     ▼                                  ▼
[ Auth & Workspace Mgmt ]      [ Groq LLM Pipeline ]
     │                            │
     │                   ┌────────┴──────────┐
     │                   ▼                   ▼
     │            [ Task Extractor ]  [ Review Queue ]
     │                   │
     └───────────────────┘
                 ▼
       [ PostgreSQL Database ]
  (workspaces, projects, users,
   tasks, assignments, statuses)
                 │
                 ▼
       [ Project Dashboard UI ]
  (Kanban view · Status · Due dates)

Tech Stack & Rationale

●React Native + JavaScript: Cross-platform mobile app (iOS & Android) with native feel

●FastAPI + Python: Async high-throughput backend with Pydantic contracts

●Groq LLM: Ultra-low-latency inference for natural language → structured task extraction

●PostgreSQL: Relational data model: workspaces → projects → tasks → assignments

●MVP Feature: Sign up/Login: Authentication and per-workspace access control

●MVP Feature: Full Task Lifecycle: Paste text → AI extract → review → assign → due date → status → dashboard

Key Retrospectives & What I Learned

The difference between a useful AI tool and a gimmick is the review step. Relay never creates tasks silently — it always shows humans what it thinks it heard and lets them adjust before anything lands in the project. That single loop turns 'this AI is annoying' into 'this AI just saved me 20 minutes of copy-pasting.'