Great discussions today at the Agent Conference during our session on “Bottlenecks of AI Agent Engineering Workflow.” I had the chance to host a conversation with Karun Appapogu (Vanguard) and Rama Krishna Raju Samantapudi (ServiceNow) on pain points across the agent lifecycle—from design and implementation to deployment and iteration. One theme came up repeatedly: the hardest part isn’t building agents, it’s making them reliable. Getting a demo to work is relatively easy, but making it reliable is much harder, especially when it comes to understanding behavior across real-world scenarios, debugging failures, and iterating with confidence at scale. This is exactly the problem we’re focused on at Arklex, knowing how systems perform, where they fail, and how to improve them. 👉 Try ArkSim (open-source AI agent testing tool): github.com/arklexai/arksim #AgenticAI #AIEvaluation #AIEngineering #Arklex #AIConference #OpenSource

I wrote this blog post to share practical tips on how academics can collaborate with industry to explore alternative funding sources. Amid all the government cuts, I hope this can help other faculties. Feel free to reach out to me if you need more help. The universe conspires to help the dreamers! Academic dreamers, let the universe see your awesome work! docs.google.com/document/d/1…

Check out our collaboration with @Walmart at NAACL 2025 Hall 3, 11 am MDT today, April 30th. @ArklexAI @CUSEAS