Kelvin Amoaba
Software engineer building scalable systems and exploring the depths of low-level architecture. Currently at Vela Partners.
CoFEE: Reasoning Control for LLM-Based Feature Discovery
A framework for automating feature discovery from unstructured data using LLMs with cognitive constraints, inducing reasoning behaviors like backward chaining from outcomes and verification against data leakage criteria. Achieves 15.2% higher success rate, 29% fewer features generated, and 53.3% cost reduction over vanilla LLM approaches.
arXiv 2026
VCBench: Benchmarking LLMs in Venture Capital
The first benchmark for predicting founder success in venture capital, providing 9,000 anonymized founder profiles. State-of-the-art LLMs like DeepSeek-V3 deliver over 6x baseline precision, with most models surpassing human benchmarks.
arXiv 2025
From Limited Data to Rare-event Prediction: LLM-powered Feature Engineering and Multi-model Learning in Venture Capital
A framework for predicting rare, high-impact outcomes by integrating large language models with a multi-model machine learning architecture for venture capital decision-making.
arXiv 2025
AI-powered analyst agents for VCs. Built autonomous LLM-based agents to automate early-stage tasks like pitch deck analysis, market research, and investment memo generation—drastically reducing due diligence time.
A full-stack self-healing web scraper that leverages AI to automate complex web scraping workflows—over 10 million records collected, with a 90% reduction in manual data collection.