Inspiration Go-to-market strategy is often based on guesswork, leading to risky and expensive product launches. We wanted to create a sandbox where product managers could simulate strategies against their actual user research, de-risking their decisions with AI.
What it does ProductSIM is a visual canvas that turns GTM planning into an interactive simulation.
Upload Personas: Users upload their existing customer research documents.
Model a Strategy: On the canvas, the user types a strategy, like "Launch a PLG freemium model."
AI Simulation: Our AI agent analyzes the strategy against each persona, generating projected KPIs (CAC, conversion rate, etc.) and qualitative commentary on how they would react.
Visualize & Compare: The results appear as new nodes on the canvas, allowing for easy, side-by-side comparison of different strategic branches.
How we built it Backend: An asynchronous FastAPI backend powers the core logic.
AI Core: We used LlamaCloud to build a RAG pipeline that lets the AI reason over user-uploaded documents. The simulation is a multi-step "Plan-and-Execute" agent that first generates questions about each persona and then synthesizes the retrieved answers into a final analysis.
Frontend: A React Flow canvas provides the interactive, node-based user experience.
Database: PostgreSQL stores all user data, including the full reasoning trace of each simulation in a JSON field for context-aware follow-ups.
Challenges & Accomplishments Our main challenge was making follow-up conversations context-aware. We solved this by creating a special simulation message type that stores the AI's full reasoning trace in its metadata. When a user branches from a simulation, this rich metadata is used to construct a highly contextual prompt, allowing the AI to build on its previous analysis instead of starting over.
What's next for ProductSIM Our next step is to ground our simulations in the real world by integrating external market data using a tool like Bright Data, allowing the AI to reference competitive benchmarks and market trends.
Built With
- amazon-web-services
- anthropic
- brightdata
- fastapi
- llamacloud
- llamaindex
- next.js
- openai-(for-llm)
- python
- react-core-libraries:-llamaindex
- react-flow
- sqlalchemy-cloud-services-&-apis:-llamacloud
- typescript
- typescript-frameworks:-fastapi
Log in or sign up for Devpost to join the conversation.