CascadeIQ

Inspiration

Construction projects in India are massive crores of rupees, thousands of workers, months of planning. Yet one delayed steel delivery can trigger a domino effect that costs lakhs. We saw project managers manually tracking dependencies in spreadsheets, discovering delays only when it was too late. We built CascadeIQ to give them the power to see tomorrow's problems today.

What it does

CascadeIQ is a continuous project intelligence platform that watches your construction schedule, simulates the future, and alerts you before delays become losses. Upload your project files (drawings, schedules, supplier emails). Watch as CascadeIQ builds a live dependency graph. See how one delay triggers a cascade through your project. Compare three recovery strategies side-by-side. Save money with AI-powered recommendations. It's like Git for construction decisions - every dependency, every revision, every simulation is traceable and replayable.

How we built it

Backend: FastAPI + Python with networkx for dependency graphs, OR-Tools for optimization, and PostgreSQL for storage. We used Docling + PyMuPDF for document parsing.

Frontend: Next.js 14 + React with React Flow for beautiful graph visualization, Framer Motion for cascading animations, and Tailwind CSS for a professional UI.

AI: LightGBM for delay prediction and OpenAI/Gemini API for natural language explanations (we never let the LLM make decisions- it only explains them).

Challenges we ran into

Graph complexity - Building a dependency graph that accurately represents construction workflows was harder than expected. We solved it by creating a "Construction Genome" - JSON templates for common workflows like concrete pouring and steel erection. Performance - Simulating cascading delays across 50+ tasks needed to happen in real-time. We optimized using networkx's BFS traversal and added Monte Carlo confidence intervals for uncertainty. The WOW moment - We knew our cascade animation was the key to winning. Making red nodes propagate through the graph smoothly in under 2 seconds took hours of Framer Motion tweaking. Scope creep - We had to resist the urge to add BIM, IoT, and Digital Twin features. Instead, we focused on making our 5 core features flawless.

Accomplishments that we're proud of

Our pipeline works from upload to recommendation in under 60 seconds The cascade animation - Watching red propagate through dependencies gives judges that "whoa" moment Quantified impact - We don't just say "it helps" - we show ₹11.8 lakh savings with 88% confidence

What we learned

LLMs are for explaining, not deciding - A deterministic graph + OR-Tools optimization can be way more credible than a black-box AI

What's next for CascadeIQ

Version 2 - Live BIM integration for real-time drawing updates Version 3 - IoT sensor streaming for live site intelligence Version 4 - Autonomous procurement with smart contracts Version 5 - Full digital twin with AR/VR visualization We built CascadeIQ to win this hackathon. But we designed it to change how construction projects are managed - one delay at a time.

Built With

• api
• css
• docling
• fastapi
• flow
• framer
• motion
• networkx
• next.js
• openai/gemini
• or-tools
• postgresql
• pymupdf
• python
• react
• redis
• shadcn/ui
• sqlalchemy
• tailwind
• typescript