Vikram is both a close friend & approaching AI style products in exactly the right way. Really interesting to follow the evolution of AI support engineers as models get better at long context and more agentic. Sometimes sounds like a boring problem, but still super tricky.
Vikram Sreekanti
Vikram Sreekanti29.7.2025
Super excited to launch @RunLLM v2 today! 😎🎉 When we started building expert systems on complex products 2 years ago, we had no idea it would turn into an AI Support Engineer. We started by focusing on delivering the highest-quality answers possible, and that earned us the trust of teams at Databricks, Sourcegraph, Monte Carlo Data, and Corelight. But high quality answers are only one part of a good AI Support Engineer. Since our last launch, we’ve re-built RunLLM from the ground up, introducing a new agentic planner, redesigning our UI, and giving our customers fine-grained control over agent workflows. Today, we’re focused on the new agentic features, because, well, it’s 2025 and you have to have agents. 😉 Jokes aside, the post below outlines in full detail how we’ve rethought RunLLM’s core inference process, how we show our work, and what we’ve built to do log & telemetry analysis. Check it out! 👇
5,98K