1B+ Daily bid requests
~$140M Enterprise exit value
250M User identity graph

The Challenge

Processing massive volumes of continuous real-time bids with strict latency limits before modern cloud scaling paradigms existed, while protecting privacy.

The Approach

Designed a decentralized architecture where continuous ML learnings were compressed into Java objects on individual edge servers for sub-millisecond decisions. Pushed parallel bid requests to the browser via JavaScript with a cryptographic sandbox layer (patents: US 9,886,718 & US 20110173126).

Real-Time Bidding & ML Optimization Engine Real-Time Bidding & ML Optimization Engine Real-Time Bidding & ML Optimization Engine

Impact

Scaled the self-serve DSP to 1B+ daily bids. Established relationships with major clients including DoubleClick (Google), ValueClick (Epsilon), and Specific Media (Viant). Acquired by News Corp for ~$140M.