project pulse

This is a perfect structure for your Devpost submission. Based on our build and the code you’ve written for Project Pulse BD, here is the content you should use to fill out these sections:

💡 Inspiration The "Execution Gap." Many students and aspiring entrepreneurs in Bangladesh have brilliant ideas but get overwhelmed by the "how." We noticed that while AI can chat about ideas, it rarely gives a concrete, localized roadmap. We wanted to build a tool that bridges the gap between a "dream" and a "Day 1 task list," specifically tailored for the unique market of Dhaka.

🚀 What it does Project Pulse BD is an AI-powered "Board of Advisors" and Execution Engine. A user enters a business idea, and the app triggers a multi-agent reasoning chain. Four distinct AI agents (CMO, CFO, COO, and Risk Lead) analyze the idea from different perspectives—marketing, finance, operations, and compliance. It then generates:

A Feasibility Score and local market verdict.

A 7-Day Launch Plan with specific daily tasks.

Localized Intelligence (suggesting bKash for payments, Pathao for logistics, and Dhaka-specific marketing).

A dynamic Cost Breakdown Chart and a downloadable execution report.

🛠️ How we built it Frontend: Built with Streamlit for a responsive, dashboard-style UI.

AI Engine: Powered by Gemini 2.0 Flash via the Google Generative AI SDK. We used advanced prompt engineering to create "expert personas" for the agents.

Data Visualization: Used Pandas and Streamlit’s native charting to turn AI-generated JSON data into visual cost breakdowns.

Deployment: Containerized with Docker and deployed on Google Cloud Run (region: asia-south1) to ensure low-latency access for users in Bangladesh.

🚧 Challenges we ran into The biggest challenge was Contextual Grounding. Initially, the AI gave generic Western advice. We had to engineer a "Bangladesh Reality Layer" into the system prompt to ensure the CFO understood BDT currency and the COO understood the logistics of Dhaka traffic and local delivery services. We also faced challenges in parsing structured data (like scores and costs) out of conversational AI responses, which we solved using RegEx and JSON schema enforcement.

🏆 Accomplishments that we're proud of Multi-Agent Orchestration: Successfully getting four different AI personas to "debate" an idea and provide a unified report.

True Localization: The app successfully identifies local hurdles like trade licenses and suggests specific Bangladeshi platforms (Nagad, bKash, Steadfast).

Production-Ready: Moving from an empty folder to a live, containerized cloud application on GCP in just a few hours.

📚 What we learned We learned the power of Agentic Workflows. Instead of asking an AI "how to start a business," we learned that breaking the problem down into specific roles (CMO, CFO, etc.) produces much higher quality, professional-grade results. We also gained deep experience in deploying serverless containers on Google Cloud.

⏭️ What's next for Project Pulse Live Market Data: Integrating real-time pricing for logistics and ad spend in Bangladesh.

Multimodal Input: Allowing users to upload a photo of a handwritten sketch or a product and having Gemini analyze the "physical" business potential.

Voice Pitch: A feature where users can "pitch" their idea via voice, and the board gives immediate verbal feedback.