DealSight AI

Institutional-grade M&A due diligence in seconds, not weeks.

Built at the Hanwha AI Center HACathon — Building AI for Finance | April 3, 2026, San Francisco

What is DealSight AI?

DealSight AI compresses the M&A due diligence pipeline from weeks to minutes. A senior analyst spends 4+ hours reading a single Confidential Information Memorandum (CIM). DealSight AI processes the same document in under 60 seconds — producing a three-lens analysis (CPA, M&A, Legal), valuation range, risk flags, 12 actionable items, and 20 buyer questions ready for the seller meeting.

The same AI intelligence layer runs through all five stages of a deal — passing structured data forward automatically so nothing is re-entered between stages.

Demo Snap

Landing page

Stage 1 — CIM Analysis

Three-lens due diligence: CPA, M&A, and Legal simultaneously. Valuation range, risk scorecard, action items, buyer questions.

Stage 2 — Financial Screening

Pre-LOI financial screen from P&L and balance sheet. Revenue analysis, EBITDA normalization, owner compensation addbacks.

Stage 3 — LOI Builder

Auto-generated Letter of Intent anchor terms from CIM and financial data. Purchase price range, payment structure, transition terms.

Stage 4 — QoE Dashboard

Quality of Earnings analysis. EBITDA bridge waterfall, working capital analysis, AR aging, weighted buy/walk score.

Stage 5 — Agreement Review

AI-powered purchase agreement review. Closing checklist, clause flags, risk items, recommended closing terms.

The Five-Stage Pipeline

CIM Analysis → Financials → LOI Builder → Diligence → Close

Stage	Input	What AI does	Time saved
CIM Analysis	CIM PDF	Three-lens analysis: CPA, M&A, Legal. Valuation range. 20 buyer questions.	4 hrs → 60 sec
Financials	P&L / balance sheet CSV or PDF	EBITDA normalization, owner addbacks, revenue quality screen	3 hrs → 90 sec
LOI Builder	Auto — no upload	Price range, deal structure, earnout, transition terms	4 hrs → instant
QoE Dashboard	Full financial package	EBITDA bridge waterfall, working capital, AR aging, buy/walk score	3 weeks → hours
Agreement Review	APA / purchase agreement	Clause flags, closing checklist, negotiation recommendations	6 hrs → 60 sec

Tech Stack

Layer	Technology	Why
Frontend	React + TanStack Start	Lovable-generated, exported and owned
Backend	Python FastAPI	Real code, not no-code. Deployed on Railway.
AI	Claude API (claude-sonnet-4-20250514)	Reads PDFs natively. One call = three expert lenses.
Database	Supabase (Postgres)	Deal persistence across sessions
Frontend hosting	Cloudflare Workers	Matches TanStack Start deployment target
Backend hosting	Railway	Free tier, auto-deploy from GitHub

Architecture

Browser (Cloudflare Workers)
    │
    │ FormData (PDF/CSV upload)
    │ or JSON (LOI Builder)
    ▼
FastAPI Backend (Railway)
    │
    │ POST /v1/messages
    │ { model, system_prompt, document }
    ▼
Claude API
    │ Returns structured JSON
    ▼
FastAPI parses + enriches
    │ Revenue trend %, valuation midpoint,
    │ flag counts, EBITDA bridge waterfall
    ▼
JSON response → React renders dashboard
    │
    ▼
Supabase — deal stored for persistence

Data flows forward automatically

Stage 1 output → stored in React state
     ↓
Stage 2 receives step1_output (prefills CIM context)
     ↓
Stage 3 receives step1_output + step2_output (auto-generates LOI)
     ↓
Stage 4 receives step1_output + step3_output
     ↓
Stage 5 receives step1_output + step3_output

No re-entry of data between stages.

Repository Structure

Deal-Sight-AI/
├── dealsight-backend/
│   ├── main.py                        ← FastAPI backend 
│   ├── requirements.txt
│   ├── Procfile
│   ├── runtime.txt
│   └── prompts/
│       ├── step1-cim-system-prompt.txt
│       ├── step2-financials-prompt.txt
│       ├── step3-loi-prompt.txt
│       ├── step4-qoe-prompt.txt
│       └── step5-agreement-prompt.txt
├── screenshots/
│   ├── Firstpage.png
│   ├── cim-analyser.png
│   ├── finance.png
│   ├── loi-builder.png
│   ├── diligence.png
│   └── close.png
└── README.md

Frontend lives in a separate repo: dealsight-ai-frontend — React + TanStack Start, deployed on Cloudflare Workers

API Endpoints

All endpoints live at https://deal-sight-ai-production.up.railway.app

Method	Endpoint	Input	Description
`POST`	`/api/cim-analyzer`	`file` (PDF)	Three-lens CIM analysis
`POST`	`/api/financials`	`file` (PDF/CSV) + `step1_output`	Financial screening
`POST`	`/api/loi-builder`	`step1_output` + `step2_output` (JSON)	LOI anchor terms
`POST`	`/api/qoe-dashboard`	`files` (CSV/PDF, multiple) + `step1_output` + `step3_output`	QoE analysis
`POST`	`/api/agreement-review`	`file` (PDF) + `step1_output` + `step3_output`	Agreement review
`GET`	`/api/deals`	—	List all deals
`GET`	`/health`	—	Health check

Interactive API docs: https://deal-sight-ai-production.up.railway.app/docs

Setup Guide

Prerequisites

Python 3.11+
Node.js 18+
Anthropic API key
Supabase account (free tier)
Railway account (free tier)
Cloudflare account (free tier)

1. Backend setup

git clone https://github.com/KamayaniR/Deal-Sight-AI.git
cd Deal-Sight-AI/dealsight-backend

python3 -m venv venv
source venv/bin/activate
pip install -r requirements.txt

Create .env:

ANTHROPIC_API_KEY=sk-ant-...your-key...
SUPABASE_URL=https://your-project.supabase.co
SUPABASE_KEY=eyJ...your-anon-key...
ALLOWED_ORIGINS=http://localhost:8080,https://your-frontend.workers.dev

Run locally:

uvicorn main:app --reload --port 8001

Test:

curl http://localhost:8001/health
# {"status":"ok"}

curl -X POST http://localhost:8001/api/cim-analyzer \
  -F "file=@your-cim.pdf"

2. Supabase setup

In Supabase SQL Editor, run:

CREATE TABLE deals (
  id UUID DEFAULT gen_random_uuid() PRIMARY KEY,
  created_at TIMESTAMP WITH TIME ZONE DEFAULT NOW(),
  business_name TEXT,
  industry TEXT,
  stage INTEGER DEFAULT 1,
  cim_analysis JSONB,
  financials JSONB,
  loi_terms JSONB,
  qoe_analysis JSONB,
  agreement_review JSONB,
  status TEXT DEFAULT 'active'
);

3. Deploy backend to Railway

# Push to GitHub, then:
# railway.app → New Project → Deploy from GitHub
# Add environment variables: ANTHROPIC_API_KEY, SUPABASE_URL, SUPABASE_KEY, ALLOWED_ORIGINS

4. Frontend setup

git clone https://github.com/KamayaniR/dealsight-ai-frontend.git
cd dealsight-ai-frontend
npm install
npm run dev
# Open http://localhost:8080

Create .env.local:

VITE_API_URL=https://deal-sight-ai-production.up.railway.app

5. Deploy frontend to Cloudflare

Connect dealsight-ai-frontend repo to Cloudflare Workers & Pages:

Build command: npm run build
Deploy command: npx wrangler deploy
Add environment variable: VITE_API_URL

Token Limits

Endpoint	Max tokens	Why
CIM Analyzer	16,000	Large system prompt + full PDF + detailed JSON output
Financials	6,000	Shorter prompt, structured financial data
LOI Builder	8,000	Text-only, concise deal terms
QoE Dashboard	12,000	Multiple CSV files + detailed EBITDA analysis
Agreement Review	12,000	Full legal document + detailed clause flags

Cost

Item	Cost
Claude API per CIM analysis	~$0.05
Full five-stage pipeline per deal	~$0.20
$10 in API credits	~50 full pipelines
Railway (free tier)	$0/month
Supabase (free tier)	$0/month
Cloudflare Workers (free tier)	$0/month

Business Model

Per-analysis pricing:

Single CIM scan: $50
Full five-stage pipeline per deal: $500
Enterprise unlimited: $2,000/month per deal team

Target customers:

Lower middle market PE firms reviewing 200-500 CIMs per year
M&A advisory firms reducing junior analyst hours
Search fund operators accessing institutional-grade DD at startup cost

Unit economics:

API cost per full pipeline: ~$0.20
Price per full pipeline: $500
Gross margin: 99.96%

Demo Mode

Press Shift+D anywhere in the app to load hardcoded demo data for the Midwest MSA Managed IT Services deal — no API call needed, no PDF required.

Demo deal: Managed IT Services Provider, Colorado

Revenue: $2.44M | Net Margin: 33.4% | Recurring: 80%
Verdict: PROCEED WITH CONDITIONS
Valuation: $3.3M — $5.0M

Built At

Hanwha AI Center HACathon — Building AI for Finance San Francisco, April 3, 2026

Sponsors: Lovable · n8n · MiniMax · Crossmint · Hanwha

License

MIT License — see LICENSE for details.

DealSight AI — Institutional-grade M&A due diligence for everyone in the deal.

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Repository files navigation

DealSight AI

What is DealSight AI?

Demo Snap

Landing page

Stage 1 — CIM Analysis

Stage 2 — Financial Screening

Stage 3 — LOI Builder

Stage 4 — QoE Dashboard

Stage 5 — Agreement Review

The Five-Stage Pipeline

Tech Stack

Architecture

Data flows forward automatically

Repository Structure

API Endpoints

Setup Guide

Prerequisites

1. Backend setup

2. Supabase setup

3. Deploy backend to Railway

4. Frontend setup

5. Deploy frontend to Cloudflare

Token Limits

Cost

Business Model

Demo Mode

Built At

License

About

Uh oh!

Releases

Packages

Uh oh!

Contributors

Uh oh!

Languages

Name		Name	Last commit message	Last commit date
Latest commit History 23 Commits
prompts		prompts
screenshots		screenshots
.DS_Store		.DS_Store
.gitignore		.gitignore
LICENSE		LICENSE
Procfile		Procfile
README.md		README.md
main.py		main.py
requirements.txt		requirements.txt

Folders and files

Latest commit

History

Repository files navigation

DealSight AI

What is DealSight AI?

Demo Snap

Landing page

Stage 1 — CIM Analysis

Stage 2 — Financial Screening

Stage 3 — LOI Builder

Stage 4 — QoE Dashboard

Stage 5 — Agreement Review

The Five-Stage Pipeline

Tech Stack

Architecture

Data flows forward automatically

Repository Structure

API Endpoints

Setup Guide

Prerequisites

1. Backend setup

2. Supabase setup

3. Deploy backend to Railway

4. Frontend setup

5. Deploy frontend to Cloudflare

Token Limits

Cost

Business Model

Demo Mode

Built At

License

About

Resources

License

Uh oh!

Stars

Watchers

Forks

Releases

Packages 0

Uh oh!

Contributors

Uh oh!

Languages

Packages