Nitin Bhardwaj
← IndexCase study · 03 / 11

Leera.

An AI product manager for modern engineering teams.
Client
Leera AI
Role
Founding Engineer · Architecture & Platform
Year
2023
Duration
3.5 years
Stack10 tools
  • Next.js
  • TanStack
  • Zustand
  • Tailwind
  • shadcn/ui
  • Radix UI
  • Turborepo
  • React Native
  • Tauri
  • Vercel
Leera — cover
01Outcomes at a glance
  • Teams onboarded

    5,000+

  • Faster sprints

    2×

  • Less planning time

    − 60%

02The brief

Modern teams hate ticket-keeping. Existing tools (Jira, Linear, Asana) optimise for control; engineers wanted a system that did the management work for them — without sacrificing the rigour of agile.

03Approach

Three moves that defined the engagement.

  1. 01

    Designed a TipTap-native workspace where @Leera AI is a first-class citizen of every comment, description and thread.

  2. 02

    Built autonomous agents on a small, composable primitive — standup, backlog refinement, sprint planning — each running on its own schedule.

  3. 03

    Engineered deep, real-time integrations with Git, Slack, Figma and Sentry so the AI is never asking 'did you write the code?'.

04What's in the box

Every app, service and worker that ships with the product — and where each one runs.

A single Turborepo with three first-class clients — web, mobile and desktop — sharing one type system, one auth layer and one realtime engine.

Apps & clients04
  • Web

    Web app

    The canonical workspace — boards, docs, sprints and inline @Leera AI agents. The surface every team logs into.

    • Next.js
    • TanStack
    • Zustand
    • Tailwind
    • shadcn/ui
    • Radix UI
    • TipTap
  • Mobile

    Mobile app

    On-call companion for engineers — standups, ticket triage, AI summaries and push when sprints drift off-track.

    • React Native
    • Expo
    • TanStack
    • Zustand
  • Desktop

    Desktop app

    A native menubar client that surfaces priorities and AI nudges without ever pulling you out of focus mode.

    • Tauri
    • Rust
    • React
    • Tailwind
  • API

    Realtime API

    Multiplayer state, presence and document sync — feeds every client over a single coordination fabric.

    • Node
    • TypeScript
    • Postgres
    • Redis
Services & workers02
  • Service

    Agent runtime

    Schedules and runs autonomous agents on top of a small composable primitive — one shape, many specialised behaviours.

    • Node
    • TypeScript
    • BullMQ
    • Redis
  • Service

    Integrations service

    Bi-directional sync with the team's source of truth — code, chat, design and observability — so the AI never asks 'did you write the code?'.

    • Node
    • Webhooks
    • OAuth
    • Postgres

Shared packages: a single design system, a typed RPC layer, an auth client and a realtime SDK — all consumed identically by web, mobile and desktop.

05Stack & rationale

Each choice is paired with the reason it earned its slot on this project.

  1. 01Monorepo & shared core

    Three clients, one truth. Turborepo lets web, mobile and desktop import the same RPC types, design system and realtime client — no drift between platforms, one PR ships everywhere.

    • Turborepo
    • TypeScript
  2. 02Web client

    Next.js for the App Router and edge-rendering speed; TipTap as the document substrate so @Leera AI is a first-class citizen of every comment, description and thread.

    • Next.js
    • TanStack
    • Zustand
    • Tailwind
    • shadcn/ui
    • Radix UI
    • TipTap
  3. 03Mobile & desktop

    React Native for the phone (one codebase, two stores) and Tauri for the menubar — a small native binary instead of shipping a Chromium per user.

    • React Native
    • Tauri
  4. 04Realtime & data

    Boring, durable foundations under a hand-rolled coordination layer — chosen so the data model stays inspectable, migratable and auditable as the product grows.

    • Postgres
    • Redis
  5. 05AI & agents

    A small agent primitive — schedulable, observable, composable — so a team can describe a new agent in a sentence and ship it the same day. The model layer behind it is interchangeable on purpose.

    • BullMQ
  6. 06Infra

    Vercel for the web edge, AWS for the long-running services. Boring, observable, recoverable.

    • Vercel
    • AWS
06Selected screens
Leera workspace

Chat with Leera AI to create tickets, generate burndowns, and reshuffle sprints in plain English.

Leera AI chat

Priority based kanban board

Leera agents

Standup Agent, Backlog Refinement Agent, and a primitive to build any agent your team can describe.

07Outcome

Leera reached 5,000+ teams and 2 billion+ tracked notifications across the platform & sister product, with customer-reported 2× sprint velocity and 60% less planning overhead.

Next project04 / 11

PushPilot— AI push notifications, fully on autopilot.