AI Dealflow Hunter

Inspiration

This project was inspired by a simple observation in early-stage investing:
“Fundraising is broken at seed stage.”

Early-stage investing is chaotic. Investors spend countless hours scrolling through LinkedIn, X, job boards, accelerator announcements, and news sites just to figure out which startups are growing. At the same time, many great founders raising capital remain invisible because they don’t have a network, don’t get media coverage, or are just too early.

As a builder and participant in innovation challenges, I noticed that many promising startups never get discovered in time. This inspired the question:
“What if AI could hunt startup deals for us?”
This became the spark that led to AI Dealflow Hunter — an automated system that discovers high-growth startups, extracts traction signals, and presents investment insights inside a clean dashboard.

What it does

AI Dealflow Hunter is a platform that:

Automatically finds startups from the internet
Scrapes websites, pricing pages, and career pages
Detects hiring spikes, launch announcements, and momentum
Scores traction, quality, and growth signals
Generates VC-style investment memos using AI
Presents everything in a clean, investor-only dashboard

Users can filter, compare startups, and add them to watchlists.
The goal is simple: less time searching, more time investing.

How we built it

The system has three main layers:
1. Data Discovery & Scraping
Firecrawl is used to fetch site content, job listings, blog posts, and product pages. The scraper is resilient to layout differences and finds relevant links based on patterns and keywords.

2. Multi-Agent Intelligence
LangGraph orchestrates several agents:

Scraper Agent → collects raw data
Parser Agent → extracts structured information
Analyst Agent → scores traction and risk

3. Frontend & Storage
Built with Next.js, TypeScript, and Tailwind.
PostgreSQL on Supabase stores all startup data, metrics, scores, and reports.

Challenges we ran into

Scraping inconsistencies
Every website is different. Locating pricing pages, job listings, or customer logos required fallback logic and heuristics.
Incomplete signals
Most startups don’t publish metrics like MRR or customer counts. We solved this by estimating ranges based on pricing, traffic, and hiring velocity.
Coordinating multiple agents Agents needed to pass structured results downstream.
We created intermediate schemas and validation so that memo generation didn’t break.
Time pressure
Building scraping, scoring, memo generation, and a dashboard in a short timeframe required tight prioritization.

Accomplishments that we're proud of

We built an end-to-end system that goes from raw internet data to investment insights. This system demonstrated that AI can automate deal sourcing, not just document analysis.
The workflow is modular and scalable — new agents can be added easily.
Most importantly, we built something that feels genuinely useful.

What we learned

Signal is more important than scraping.
Hiring spikes, frequent releases, and product launches matter more than static text.
Multi-agent coordination is powerful.
Breaking work into separate agents with clear responsibilities made the pipeline robust.
UX wins Investors don’t want complexity.
They want clear summaries, alerts, and confident insights.
Data incompleteness is normal
Real-world data is messy. Learning how to fill gaps with estimates and reasoning was key.
Speed over perfection
Shipping a working demo is better than designing the perfect system.

What's next for AI Dealflow Hunter

We see a strong roadmap:

Email alerts for trending startups
Market shock engine to track news, competitors, and regulation
More agents for:
- tech stack detection
- founder sentiment analysis
- pricing intelligence
Investor collaboration tools
Watchlists, shared notes, co-investment chats
Deal syndication
Allow investors to reserve allocation or commit interest

Ultimately, we want to build a platform where: