InsightGridAI

We built InsightGridAI, an AI-powered business copilot that lets executives ask 'what if' questions and get instant ML-powered predictions.

Comment

InsightGridAI - DevPost Submission

Inspiration

Business leaders face critical decisions every day - expanding to new markets, opening warehouses, pricing products, managing supply chains. While traditional BI tools excel at showing historical data, we saw an opportunity: What if leaders could not just see what happened, but simulate what will happen before making million-dollar commitments?

When a manufacturing company asked us "How would a 15% fuel price increase affect our logistics costs?" we realized they needed instant answers, not 3-day SQL reports. They needed AI-powered simulations, not static dashboards. They needed predictive insights, not post-mortem analysis.

We built InsightGridAI to bridge that gap - making powerful AI-driven business intelligence as simple as asking a question.

The vision: Empower every business leader with AI-powered decision-making at their fingertips.

What It Does

InsightGridAI is an AI-powered business copilot that lets you:

  • Ask questions in plain English: "What are our top cost drivers?" → Get instant answers
  • Run simulations: "What if fuel prices increase by 15%?" → ML-powered impact analysis
  • Get business cases: "Should we open a warehouse in Toronto?" → Full ROI analysis
  • Forecast demand: "What's our revenue if demand increases by 20%?" → ML predictions

Built 100% on AWS serverless - zero infrastructure management, auto-scales, pay-per-use.

How We Built It

Tech Stack:

  • Frontend: React, TypeScript, Tailwind CSS, Recharts
  • Backend: AWS Lambda (Python 3.11), 3 serverless functions
  • Data: S3 (CSV files), AWS Glue (data catalog), Athena (SQL queries)
  • AI: Bedrock, Gemini (fast chat), Airia (deep analysis)
  • ML: Pure Python algorithms (no NumPy - Lambda-compatible)
  • Infrastructure: AWS CDK (TypeScript)

Architecture:

User → API Gateway → Lambda Functions → Athena → S3 (Your CSVs)
                                    ↓
                              Gemini/Airia AI (via BEDROCK)

Key Innovation:

  • Pure Python ML: Rewrote linear regression, seasonality detection, confidence scoring without NumPy
  • Stateless Design: No database - queries S3 directly with Athena
  • Multi-AI: Gemini for speed (2-10s), Airia for depth (60s)
  • Infrastructure as Code: Entire stack deployed with cdk deploy

Challenges We Ran Into

1. complex AI orchestration

Accomplishments We're Proud Of

  • ✅ Built production-ready platform in 48 hours
  • ✅ 100% AWS serverless architecture - no EC2, no databases
  • ✅ Pure Python ML - no external dependencies
  • ✅ Multi-AI integration - Gemini + Airia
  • ✅ Enterprise security - IAM, S3 encryption, API Gateway
  • ✅ Real-world ROI - 10x cost savings vs traditional BI
  • ✅ Zero DevOps - deployed with cdk deploy
  • ✅ Auto-scales - handles 10 to 10,000 users

What We Learned

  • AWS Serverless is powerful: Lambda + Athena + S3 can replace entire data warehouse
  • Stateless > Stateful: No database needed when S3 is source of truth
  • AI Choice matters: Different AIs for different use cases (Gemini vs Airia)
  • Pure Python ML works: NumPy not required for linear regression
  • Infrastructure as Code is essential: CDK saved us hours of manual setup
  • User experience > Technical complexity: Make it simple, make it fast

Biggest takeaway: You can build enterprise-grade platforms on AWS serverless without teams or infrastructure management.

What's Next for InsightGridAI

Short-term (3 months):

  • Real-time data streaming (AWS IoT Core)
  • Advanced ML models (LSTM for time-series, SageMaker for deep learning)
  • Multi-tenant SaaS (AWS Cognito, Amplify)

Medium-term (6-12 months):

  • AWS Bedrock integration (Claude for enterprise chat)
  • Predictive maintenance (20-30% downtime reduction)
  • Supply chain optimization (AWS Supply Chain)

Long-term Vision:

  • Improvement in Model Inferencing
  • integration of LLM to intelligently generate more charts on the analytics page
  • Also making 'niche' company-specific versions

Goal: Make AI-powered decision-making accessible to every enterprise worldwide.

Built With

Share this project:

Updates