Skip to content

SUBMISSION.md

Latest commit

 

History

History
79 lines (55 loc) · 2.92 KB

SUBMISSION.md

File metadata and controls

79 lines (55 loc) · 2.92 KB
Raw

AlphaStream India — Hackathon Submission

ET AI Hackathon 2026 — Problem Statement 6: AI for the Indian Investor

Submission Checklist

Requirement Status Location
GitHub Repository Done README.md + commit history
Architecture Document Done docs/ARCHITECTURE.md
Impact Model Done docs/impact_model.md

1. GitHub Repository

Repository: https://github.com/ridash2005/AlphaStream_Final

What judges will find:

  • Full source code: 13 AI agents, 60+ REST endpoints, React 19 frontend with 30+ components
  • README.md — project overview, quick-start setup, full API reference
  • backend/ — Python FastAPI backend (agents, connectors, pipeline, NLQ)
  • frontend/ — React 19 + TypeScript + Tailwind Bloomberg-style terminal
  • worldmonitor/ — global market intelligence module
  • docker-compose.yml — one-command deployment
  • backend/.env.example — environment variable reference

Setup: 3 commands — uv sync -> ./start.sh -> npm run dev

2. Architecture Document

Document: docs/ARCHITECTURE.md

Covers:

  • System overview — 5-layer architecture (data sources, Pathway streaming, 13-agent reasoning, DuckDB analytics, React terminal)
  • Pathway integration — Adaptive RAG with geometric retrieval, 12+ Pathway features used
  • Agent specifications — All 13 agents with input/output/technology
  • NLQ pipeline — LangGraph 7-node graph with Text2SQL, guardrails, and SSE streaming
  • Data flow — Mermaid diagram showing end-to-end pipeline
  • API layer — 60+ REST endpoints, WebSocket, SSE
  • Performance — <2s data-to-update latency, ~7s full recommendation

Supporting diagrams (in docs/):

  • system_architecture.png — full 5-layer system diagram
  • multi_agent_system.png — agent coordination and fusion
  • herd_of_knowledge.png — parallel multi-source news aggregation

3. Impact Model

Document: docs/impact_model.md

Quantified estimates with stated assumptions:

Impact Area Estimate Method
Time saved per user INR 2.19L/year 1.75 hrs/day x INR 500/hr x 250 days
Alpha generated per user INR 54,000/year 2-3 signals/month x 60% accuracy x INR 3,000
Risk reduction 15% behavioral loss reduction Signal-based early warning
ET Markets revenue (Year 1) INR 59.9 Cr/year 50K premium users x INR 999/month x 12

All assumptions explicitly stated. Back-of-envelope math in the document.

Key Numbers

  • 13 AI agents — each specializes in one analytical dimension
  • <2 seconds — news article to updated recommendation (Pathway streaming)
  • 60+ REST endpoints — production-grade API
  • 7-node LangGraph pipeline — NLQ with Text2SQL (not hallucination)
  • 5-year backtest — all signals validated with win rates
  • 14 Cr+ target users — every Indian demat account holder