The Eval Harness Behind Our Voice Agent
How we test a voice agent on real calls; score every call, calibrate the judges, and replay real failures before shipping.
Urvesh Patel / August 25, 2026
Engineering Blog
Latest engineering articles
How Eve builds reliable agents for lawyers (opens in a new tab)
How Eve built Plaintiff Bench, a gold-standard evaluation dataset of 400+ plaintiff-law tasks, to catch citation, temporal reasoning, and retrieval failures before they reach lawyers.
Eve Legal helps plaintiff law firms settle cases 60 days faster with Claude (opens in a new tab)
How Eve uses Claude across legal workflows and internal engineering systems to accelerate settlements, improve case outcomes, and cut incident triage from hours to minutes.
Allen Deng / March 19, 2026
Why we chose Turbopuffer for our search infrastructure
OpenSearch is built for a few large shards, but we needed thousands of small ones. Here’s our engineering deep dive on why we migrated to Turbopuffer.
David Zeng / August 20, 2025
Urvesh leads voice AI quality and evaluation at Eve; building the eval harness that scores every call, catches regressions before they ship, and powers the team's live experiments on the voice agent.
AI Engineer
The greatest opportunities belong to those who see where value is moving before the rest of the world does. Every industry gets rebuilt around AI exactly once — Here, we don't follow the future of law. We define it.
Member of Technical Staff
David Zeng writes about the systems and ideas behind Eve.
Co-Founder and Chief Architect
Torehan Sharman writes about the machine learning systems and ideas behind Eve.
Machine Learning Engineer
We're hiring
Help us build the systems that turn legal work from hours into minutes. Join an ambitious team working at the frontier of AI and law.
See open roles