Suraj Vathsa

Much of the valuable information in agent traces isn’t directly filterable with existing techniques like full text search or metadata filtering. AI Search in LangSmith closes that gap. You describe what you're looking for in plain natural language and can issue queries like "user frustrated" and "pii leaked" on the fly, which was previously impossible to do without defining an online evaluator ahead of time.

Really bullish on workload-aware inference! Database systems are of course a crown jewel of exploiting workload knowledge to achieve good performance. now in today's inference era, any system with a declarative interface and budding demand for LLM inference is probably a good candidate for workload-aware inference

Stoked to launch 𝚜𝚖𝚒𝚝𝚑𝚝𝚞𝚗𝚎 in public beta! Post-training your agent's policy has never been easier, with training APIs from @baseten and @FireworksAI_HQ becoming increasingly accessible. But sourcing high-quality data to use them effectively remains a challenge. LangSmith processes hundreds of millions of traces a day, making it a rich system of record for your agent’s behavior. 𝚜𝚖𝚒𝚝𝚑𝚝𝚞𝚗𝚎 turns those trajectories into the data you need for supervised fine-tuning, with an end-to-end pipeline for curation, preparation, training and deployment for models you want to SFT for specific tasks. We care deeply about giving teams ways to own their own intelligence, and as the open model frontier continues to advance, it’s becoming increasingly clear that post-training on your own agent data will become a key part of that. Excited to get this into people's hands!

Many people still confuse Chinese model with Chinese inference: - A model is like a recipe for a dish—e.g. orange chicken - Claude keeps its recipe secret, charges the dish $50. You must eat in front of Dario and he will watch you the entire time. You can’t change the ingredients or cook it privately. - Now China releases a free recipe that tastes just as good. You can use American chicken and cook it in your own kitchen (self-hosting), or have a U.S. cloud like AWS or Nebius cook it and pay $2-5. - Chinese inference is DoorDashing the finished dish from Beijing. Your order and data go there, and you must eat Beijing chicken. It’s fair to worry about the safety of their ingredients. - Regulate the kitchen, ingredients, and data. Don’t ban the recipe.

Own the frontier: "In our evals, Nemotron 3 Ultra with a tuned LangChain Deep Agents harness provides advanced agent performance at a much lower inference cost. The main takeaway is that agent performance improves when the model, harness, evals, and runtime are tuned together."