Consultants who build agents that work! Building AI is easy, but building Agents that integrate with your existing system is hard.
Chicago, Il
Joined February 2019
- Tweets130
- Following16
- Followers300
- Likes57
Scaling agent development across a regulated firm means solving for access, reliability, and control.
Inside our work with Stifel: governed Claude access, a shared MCP gateway, and LangSmith evals in release pipelines.
focused.io/case-studies/gove…
Great senior engineers don’t always have agent experience. We’re investing in helping them build it.
Introducing the Focused Agent Academy: 12 weeks of paid, hands-on training and real client work.
Starts Nov 2 in Denver & Chicago. Apply: focused.io/agent-academy
NYC agent builders: tomorrow night is the Agentic Lounge at Fraunces Tavern, ahead of LangChain Interrupt.
Talk production agents with Focused’s principal engineers over cocktails, bites, and live piano.
No Interrupt ticket needed: luma.com/7vezdbxt
A piano bar full of people who build agents.
Join us Sept 23 at Fraunces Tavern in NYC for conversations with Focused’s principal engineers, cocktails, food, and live piano. No Interrupt ticket needed.
Haven’t grabbed your spot yet? luma.com/7vezdbxt
Focused retweeted
All my NYC friends we are hosting an Agentic Lounge (fancy!) the night before @LangChain Interrupt on Sept 23rd.
Come hang, win tickets to Interrupt, win other stuff. Have fun talk to nerds
luma.com/7vezdbxt
Next week in Chicago, Jeremy Smolens is breaking down how MCP Apps + @LangChain enable dynamic, task-aware UIs.
Join us: luma.com/1oxz1env
#LangChain
NYC, we’re starting @LangChain #Interrupt week at the Focused Agentic Lounge!
Come for themed cocktails, food, live music, and a room full of agent builders.
Bring the agent keeping you up at night and talk it through with Focused’s senior + principal engineers.
Sept 23: luma.com/7vezdbxt
Focused retweeted
🚀Today we launched LangSmith Tuned Evaluators, starting with Perceived Error.
Tuned Evaluators run on production traces to catch undesirable agent behavior and attach feedback that you can use in your agent improvement processes.
In our benchmark, our tuned model beat frontier models at 82% lower cost.
langchain.com/blog/introduci…
Focused Lead Agent Engineer Emilio Navarro was chatting with fellow engineers about which LLM was the best at structured data extraction.
So naturally, we ran an experiment in LangSmith.
Full experiment + repo here:
focused.io/lab/benchmarking-…
Focused retweeted
We’re making LangSmith Sandboxes available to try for free.
Start today: smith.langchain.com/
Focused retweeted
The EU AI Act is the first comprehensive regulation for AI systems, and it’s right around the corner.
With the clock running down, we took a deeper look at what these requirements mean for your agents at the Article level, and how LangSmith + @LangChain_OSS directly address them.
Good stuff. Thanks for sharing @LangChain
LangChain Community Spotlight: Distributed LangGraph Architecture with RemoteGraph 🔗
Production-ready example for distributed LangGraph apps with HTTP-invoked subgraphs. Key insight: RemoteGraph returns full state, not deltas—parents handle reducers.
From Focused.io.
📦 github.com/focused-dot-io/re…
Only three weeks until Interrupt!
As a LangChain partner, we help teams get agents integrated, observed, and running reliably inside the systems they already have.
If you're building and need agent expertise, grab time with us at Interrupt. See you there!
focused.io/langchain/interru…
Spent last week at #O11yDayNYC alongside our partners at Honeycomb.
Much of the conversation focused on how observability evolves as systems become more complex. Exactly the work we’ve been leaning into together.
Big thanks to the Honeycomb team for bringing everyone together for such a great event.
Full takeaways ↓
focused.io/lab/what-we-learn…
Our newest use case is up! Streaming agent state with LangGraph.
If your agent takes ~9 seconds to run a workflow, don’t make users stare at a blank screen.
Stream node updates, tool calls, and LLM tokens to the UI while it runs.
focused.io/lab/streaming-age…
Focused x LangChain: Detecting Errors Early with LangSmith xitter.cf/i/broadcasts/1pKkOyXWL…
Denver builders 👋
Tomorrow: Official Denver LangChain Meetup.
We’re breaking down what it actually takes to run LangChain systems in production.
Real incident walkthrough. Real lessons learned.
If you’re building agents or moving from experiments to production, join us.
Food + drinks provided.
luma.com/x8kffsta
We’re hosting the first official LangChain meetup in Denver in just two weeks!
Lead Agent Engineer Emilio Navarro will walk through how to detect errors early with LangSmith and why eval and observability are critical in production agent systems.
We hope you'll join us → luma.com/x8kffsta
If you have a good answer you get an interview
I have an interview question I was told is too hard for the Agent Engineer position. How would you solve it?
If you do well you can skip the first two interviews.
Client has asked us to create a RAG for discovery documents for trial. Documents are transcripts from depositions, emails, and SMS.
The client would like the AI to answer the question "has X person ever contradicted themselves, show me where?"
Explain how you'd create a agentic RAG that can solve this problem.