Curiosity killed the cat, but satisfaction brought it back.
Next →The best way to reach me is a message on LinkedIn.
LinkedInBut I also post more casually on X.
@jess_weng18Beyond Simple Agents: Building with LangChain Deep Agents
A lightning lesson with The Gen Academy. Covers what sets Deep Agents apart, how the harness handles long-running work, managing context with memory and summarization, and delegating to subagents.
View on MavenHow Deep Agents Remember Without Keeping Everything in Context
In this talk I covered 4 context engineering techniques with the deep agents harness (summarization, subagents, offloading and long-term memory) and how a model-agnostic harness keeps memory intact across models.
Event page on O’ReillyAn introduction to agent harnesses and Deep Agents; covering the system prompt, tools, human-in-the-loop and subagents.
+ a live demo building a journaling agent (mental health is important).
Watch the recordingA free LangChain Academy course on building long-running agents with Deep Agents, LangChain’s open-source agent harness: covering sandboxes, context management, subagents and more.
Take the course Read the launch postA free LangChain Academy course on building no-code agents in LangSmith Fleet: covering human-in-the-loop review, MCP, subagents and so much more.
Take the course Read the launch postEducation Engineer
Helping developers learn and build with hands-on content across LangChain’s ecosystem.
M.S. Business Analytics
B.A. Computer Science
Scroll up to climb
The roots: early experience that helped the beanstalk grow.
Partnerships & Strategy
via Align Communications & Creative
May – Aug 2025Consumer Insights
Jan – May 2025Data Analytics
Jan – May 2025Quick explainers, five minutes or less.
Everything around the model that turns it into a work engine.
Folders of instructions an agent only opens when it needs them, so its context stays lean.
LangChain’s open-source harness for long-running work: context engineering and multi-agent orchestration.
How Stripe built Kai, a company-wide, always-on AI assistant.
Longer reads on building agents.
Long-running agents degrade as context piles up. How offloading, summarization, skills, subagents and memory keep an agent’s context window focused.
Read the postPhotos, memes and moments along the way
Proof I am not AI, despite the accusations :p
