Vivek Trivedy
Bio
Vivek leads Applied Research at LangChain Labs focused on making Continual Learning accessible to the world's agents. Before that he led work on Deep Agents, LangChain's open source agent harness. He's obsessed with all things agent improvement and harnesses. And also computer vision from a previous PhD life.
Session (1)
Vivek Trivedy leads open-source agents and harnesses work at LangChain and is a leading voice on the "agent harness" paradigm — Agent = Model + Harness — showing through LangChain's blog series how harness engineering alone (self-verification loops, environmental context, reasoning-compute allocation) can lift a coding agent's Terminal Bench 2.0 ranking without changing the underlying model. His session offers a concrete, practitioner's view of how production agentic systems are built, evaluated with evals, and iteratively improved at one of the field's most influential labs.
Recent talks (2)
GitHub
@vtrivedyRecent writing (6)
- Better Harness: A Recipe for Harness Hill-Climbing with Evals · blog · Apr 2026
- How we build evals for Deep Agents · blog · Mar 2026
- The Anatomy of an Agent Harness · blog · Mar 2026
- Improving Deep Agents with harness engineering · blog · Feb 2026
- Agents Should Be More Opinionated · blog · Nov 2025
- The Claude Code SDK and the Birth of HaaS (Harness as a Service) · blog · Sep 2025
Podcasts & interviews (1)
- The Never Ending Lore of Harness | Vivek Trivedy (Product Lead, Langchain) · Ground Zero AI Talks · May 2026
Public activity researched automatically · as of Jun 2026