How Zip Uses LangSmith and LangGraph to Ship New Features Faster

Dexter Lee, Staff Machine Learning Engineer at Zip, the procurement platform, explains what changed when his team started using LangGraph and LangSmith for its AI development. Before, Zip built its own code for calling LLMs, piping and storing results, and building context: adding a feature to an early version of its chatbot or agent took weeks, and its in-house traces pipeline was missing details the team needed. With LangGraph and LangSmith, traces of how nodes interact and how LLM calls happen were there without adding a line of code, other teams with their own pipelines quickly adopted LangGraph and LangSmith, and Zip built an initial evaluation setup to make sure it isn't pushing regressions. Chapters: 0:00 Zip and the LLM code it built from scratch 0:21 Slow iteration and missing traces 0:49 The first aha moment with LangGraph and LangSmith 1:16 Other teams adopt LangGraph and LangSmith 1:27 Evaluations that catch regressions 1:40 Why building it yourself sets you back Resources: → LangGraph: https://www.langchain.com/langgraph → LangSmith: https://www.langchain.com/langsmith → Intro to LangSmith course on LangChain Academy: https://academy.langchain.com/courses/intro-to-langsmith