VC StartUp Survival

Machine Learning Generative AI MLflow Streamlit LightGBM XGBoost

System Architecture

An end-to-end ML system for predicting startup survival and success using structured venture capital data. Multiple data sources — WikiData and Wikipedia APIs — are integrated to construct a rich dataset capturing funding patterns, founder attributes, and industry characteristics.

A scalable Bronze → Silver → Gold data pipeline underpins the system. Bronze ingests and augments raw data; Silver normalises and models it into a star schema (FACT_STARTUP connected to DIM_INDUSTRY, DIM_FUNDING, DIM_FOUNDERS); Gold trains, evaluates, and tracks models using MLflow. Derived features — funding gaps, founder experience, temporal investment trends — are engineered to boost predictive performance.

Result: A robust, scalable ML pipeline capable of handling real-world VC data, with LightGBM, XGBoost, and GRU models tracked via MLflow and served through a live Streamlit app.

Built with

Python LightGBM XGBoost PyTorch GRU MLflow Streamlit WikiData API

Bronze→Silver→Gold pipeline from WikiData/Wikipedia APIs through star-schema modelling; three model families tracked in MLflow, served via live Streamlit app.

← Emotion Regression w/ MoE Next: AniOntology AI →