ProdPlex

Inspiration

Product managers spend too much time buried in documentation, research, and coordination — not actual strategy. We wanted to change that. Inspired by the PNC challenge “Reimagining the Life of a Product Manager with GenAI”, we built ProdPlex, an AI-powered multi-agent copilot that automates the tedious parts of product management from research to design to analytics all using NVIDIA’s Nemotron models.

What it does

Prodplex acts like a team of intelligent assistants, each powered by a different Nemotron model. Users can seek help with a specific aspect like prototyping, market reseaqrch, strategy, etc. or combine all the agents to form an effective PRD in minutes.

How we built it

Javascript, Python, Nemotron, React.js.

Challenges we ran into

Coordinating multiple Nemotron models.

Managing prompt consistency between small and medium Nemotron models.

Integrating visual (Figma) and textual (Nemotron) systems into one workflow.

Accomplishments that we're proud of

Successfully built a multi-agent system without a central orchestrator.

Incorporated RAG into a Nemotron model for improved accuracy and research validity.

Demonstrated collaborative intelligence between Nemotron models.

Integrated real product design generation (Figma) with AI-generated PRDs.

Created a modular, low-cost system that scales easily based on user needs.

What we learned

Smaller Nemotron models can deliver great results when scoped to specific domains.

Shared-context design promotes emergent collaboration.

Prompt structure and data caching are key to stability and speed.

PM tasks are ideal for AI modularization because they naturally divide into clear roles.

What's next for ProdPlex

Add Notion/Linear integrations for real-time PM syncing.

Expand to voice-driven task updates for mobile PMs.

Introduce learning memory for ongoing product evolution tracking.

Built With

• javascript
• nemotron
• react