Automating Venture Capital: Founder assessment using LLM-powered segmentation, feature engineering and automated labeling techniques

TL;DR

This paper addresses venture capital decision-making by predicting startup founder success from founder-centric data using an integrated ML/LLM framework. It employs chain-of-thought prompting and few-shot LLM techniques to generate education-derived features, founder summaries, ten-level and twenty-persona segmentations, and 23 boolean flags, which are then evaluated with linear regression, random forest, and XGBoost on constrained datasets. Results show meaningful relationships between higher level categories and success, as well as distinctive patterns across personas and boolean flags, while model performance depends on data balance and the chosen metric, with precision being crucial for VC portfolios. The approach demonstrates a scalable path for augmenting investment decisions with AI-driven founder profiling, while acknowledging limitations such as the independence assumption for solo founders and potential LLM biases that warrant guardrails and further research.

Abstract

This study explores the application of large language models (LLMs) in venture capital (VC) decision-making, focusing on predicting startup success based on founder characteristics. We utilize LLM prompting techniques, like chain-of-thought, to generate features from limited data, then extract insights through statistics and machine learning. Our results reveal potential relationships between certain founder characteristics and success, as well as demonstrate the effectiveness of these characteristics in prediction. This framework for integrating ML techniques and LLMs has vast potential for improving startup success prediction, with important implications for VC firms seeking to optimize their investment strategies.