PM-GPT is an AI-powered Product Management Copilot that transforms ambiguous product challenges into structured problem insights, prioritized decisions, and defensible product roadmaps.
This project is intentionally designed to reflect real-world product management thinking, not surface-level automation. Each output mirrors how experienced Product Managers frame problems, evaluate trade-offs, and communicate decisions clearly to stakeholders and leadership.
Note: PM-GPT was built as a flagship portfolio project to demonstrate real-world PM decision-making, not as a tutorial, demo toy, or generic AI application.
🔗 Live Demo: https://pm-gpt-copilot.streamlit.app
Alongside the codebase, I maintain a public product journal documenting the thinking, decisions, trade-offs, and reasoning behind building PM-GPT.
This journal focuses on:
- Product thinking under ambiguity
- Framework-driven prioritization decisions
- Design trade-offs and scope boundaries
- Intentional, long-term product development
These entries document how and why PM-GPT makes decisions — not just what it outputs.
-
Journal Entry 1 — Building a Product That Teaches Product Thinking
Explores ambiguity as the core challenge in Product Management and the vision behind PM-GPT Copilot.
🔗 https://medium.com/@urishita8888/pm-gpt-copilot-day-1-bf09b75486bb -
Journal Entry 2 — Choosing the Right Framework Under Ambiguity
Examines why framework selection is a product decision, how misapplied frameworks lead to poor outcomes, and how PM-GPT reasons intentionally under uncertainty.
🔗 https://medium.com/@urishita8888/choosing-the-right-framework-under-ambiguity-c758bc7c03dc -
Journal Entry 3 — Product Decisions Are Trade-offs, Not Choices
Explores how real product decisions are shaped by constraints, opportunity costs, and the discipline of saying no, emphasizing judgment over certainty in PM work.
🔗 https://medium.com/@urishita8888/product-decisions-are-trade-offs-not-choices-6e138ff89a1fThis project and journal are actively evolving as PM-GPT grows in scope and depth.
A visual walkthrough of PM-GPT’s end-to-end decision reasoning flow:
A short walkthrough demonstrating PM-GPT in action is available in the repository:
📂 demo/pm-gpt-demo.webm
(This may later be replaced with a hosted video link.)
- Identifies the core problem behind vague or noisy product signals
- Surfaces where user value and business outcomes are at risk
- Frames problems using impact, urgency, and real-world constraints
- Generates realistic, high-leverage solution ideas
- Aligns features with problem archetypes and business context
- Avoids speculative or impractical product thinking
- Automatically selects the most appropriate framework (RICE, ICE, Kano, MoSCoW)
- Clearly explains why a framework fits the situation
- Enables side-by-side comparison of prioritization outcomes
- Simulates executive and stakeholder pushback
- Documents trade-offs, risks, and consciously rejected alternatives
- Produces PM-grade narratives suitable for reviews and interviews
- Builds a realistic 6-month product roadmap
- Balances strategy, execution, and resource constraints
- Exports a professional PDF with full decision reasoning
- Python 3.9+ (tested on Python 3.11)
- pip
git clone https://github.com/nagarishitaupputuri2007/pm-gpt.git
cd pm-gpt
pip install -r requirements.txt
streamlit run ui/app.pypm-gpt/
├── product/ # Core PM reasoning and decision logic
├── roadmap/ # Roadmap generation & PDF export
├── ui/ # Streamlit user interface
├── tests/ # Test suites
├── assets/ # Screenshots and demo media
├── demo/ # Product demo recordings
└── README.md
- Aspiring Product Managers (APM / PM / Intern roles)
- PM interview preparation and case discussion practice
- Recruiters evaluating real-world product decision-making ability
- Anyone interested in structured, explainable PM thinking
This project prioritizes clarity, explainability, and reasoning quality — the same standards expected in real-world PM teams and BigTech product organizations.
Upputuri Naga Rishita
B.Tech CSE (AI & Future Technologies)
SRM University AP
This project is licensed under the MIT License.