Introducing VenturePunks: AI-Matched Investors for Your Portfolio

After weeks of coming across job postings for a hands-on CTO to build an investor to startup matching platform, I decided to build one myself.

Venturepunks is a privacy-centric, white label matchmaking platform intended for accelerator and incubator program managers.

I have spent decades working with and for startups, fundraising, investor due diligence, M&A. I have been a founder, an investor, a mentor and an employee - and recently launched methodpunks.com to help entrepreneurs and founders and the incubators and accelerators that power them.

Matchmaking is a well-solved problem - just ask my former colleagues at Demonware where we scaled matchmaking to 600M Call of Duty players.

Add a tiny fraction of compute from an LLM and a sprinkling of some fairly basic machine learning algorithms - et voila: venturepunks.ai.

The Matching Engine

The matching engine scores every investor in your database against a given startup using four independent signals, then combines them into a single ensemble score (0-100).

Signal 1: Sector Fit

Measures overlap between the startup's sectors and the investor's sector focus using Jaccard similarity. A fintech startup matched against an investor focused on fintech + SaaS scores high; against a biotech-only investor, it scores zero.

Signal 2: Stage Alignment

Binary check: does the investor's stage preference include the startup's current stage? A seed-stage startup matched against an investor who writes seed + Series A checks scores 100%. Against a growth-only investor, it scores 0%.

Signal 3: Thesis Compatibility

Semantic similarity between the investor's investment thesis and the startup's description. Uses a sentence-transformer embedding model to generate vector representations, then computes cosine similarity. This captures nuance that simple tag matching misses - an investor whose thesis mentions "enterprise workflow automation" will score well against a startup building "B2B process orchestration tools" even if their sector tags don't overlap perfectly.

Signal 4: Network Proximity

Graph-based scoring that searches the relationship network for warm intro paths between the startup's founder and the investor. Uses a recursive shortest-path algorithm across the network graph. A 2-hop connection (founder → shared board member → investor) scores higher than a 4-hop path, and much higher than no path at all. This signal only activates when network data exists.

The four signals are then combined with adaptive weights into a single ensemble score.

Who Is VenturePunks For?

• Admins: Program directors, fund managers - people who run the workspace.
• Founders: Startup founders in the portfolio. They manage their company's fundraising.
• Mentors: Advisors, EIRs, board members - people who guide startups and have industry connections.
• Viewers: Board observers, limited partners (LPs), external advisors - people who need visibility but not edit access.

Demo coming soon. Register interest today @ venturepunks.ai.