Device Fingerprinting & Identity Resolution, Shipped to 1M+ Users

A production device-fingerprinting and identity-resolution system for fraud prevention at a cybersecurity startup (raised $4M at seed), shipped to 1M+ users. It re-identifies devices even as attributes like OS, fonts, timezone, or IP change over time, backed by TLS / network fingerprinting (JA4+) and multi-layered VM / emulator detection.

Highlights

• Architected and built the core engine in Go with a PostgreSQL schema for identity resolution and long-term re-identification.
• Applied Bayesian probabilistic matching to keep identities “sticky”: devices stay linked even when signals drift, by weighing evidence across signals rather than requiring exact matches.
• Built a ground-truth pipeline to validate signal stability and filtered features only where unreliability was proven.