Case Study

PM Agent

An end-to-end AI tool that eliminates the manual overhead of post-meeting documentation — automating agile workflows from transcription to sprint planning.

AI Meeting SummarizerAgile AutomationFull-Stack
95%Reduction in manual documentation time
~80%Smaller audio payloads via in-browser processing
100MBMax file upload supported
3Artifact types generated per meeting
The Problem

Meetings generate action — but documentation kills momentum.

Product Managers spend hours after every meeting manually converting discussions into PRDs, user stories, and sprint plans. This repetitive overhead delays execution, introduces human error, and creates bottlenecks across the entire product development cycle.

The Solution

An AI agent that turns conversations into structured artifacts — instantly.

PM Agent autonomously joins meetings, transcribes audio in real-time, and synthesizes the conversation into polished, editable agile documents. From recording to sprint plan, the entire pipeline is automated.

The Application

Tech Stack

Frontend

  • React (Vite)
  • TypeScript
  • TailwindCSS

Backend

  • Next.js
  • Supabase (PostgreSQL)
  • Row Level Security

AI / ML

  • Google Gemini
  • Groq
  • Whisper

APIs & Media

  • Recall API
  • Web Audio API
  • html2pdf.js

How It Works

01

Intelligent Ingestion & Transcription

Seamlessly captures meeting audio through multiple channels and converts it into structured text.

  • Deploys virtual bots via the Recall API to autonomously join Google Meet/Zoom calls, record, and extract transcripts.
  • Supports direct audio/video uploads up to 100MB for maximum flexibility.
  • Utilizes the Web Audio API to chunk and downsample audio to 16kHz Mono in-browser, reducing payload sizes by up to 80%.
Summarizer — meeting scheduling and calendar integration
02

Autonomous PM Artifacts

Transforms raw meeting transcripts into structured, actionable agile documentation.

  • Analyzes meeting context to construct detailed PRDs outlining Objectives, Scope, and core requirements.
  • Converts feature discussions into standard agile user stories with acceptance criteria.
  • Distills action items into actionable sprint backlogs with estimated sizing.
PM Dashboard — PRD Generator, User Stories, Sprint Planner tools
03

Resilient Architecture

Enterprise-grade data layer built for security, consistency, and fault tolerance.

  • Engineered a robust database layer using Supabase with complex relational data kept perfectly synced.
  • Strict Row Level Security (RLS) policies tied to Google Auth for proprietary data protection.
  • Resilient query logic with fallback between user_id and user_email identifiers for schema integrity.
Settings — account profile, bot config, and AI engine setup
04

Premium UX & Export

Polished interface with powerful output capabilities for professional delivery.

  • Modern UI with micro-animations and ambient design language.
  • In-browser markdown editing for rapid refinement of AI-generated content.
  • One-click styled PDF exports combining html2pdf.js and react-markdown for branded reports.
3.0 Agent — AI chatbot with conversation history

Architecture Flow

01Meeting AudioRecall API / Upload
02Audio ProcessingWeb Audio API — 16kHz Mono
03TranscriptionWhisper via Groq
04AI SynthesisGoogle Gemini
05Artifact OutputPRD / Stories / Sprint Plan