[In-Person] Founder Spotlight: Self-Improving Agents and Practical Guide to Building an Evaluation-to-Improvement Loop with Orkhan (SIR P26)

​Quality is the #1 barrier to putting AI agents in production, cited by 32% of teams in the latest LangChain State of Agent Engineering report, ahead of cost, latency, and security. This workshop will give the latest update on the evaluation and optimization frameworks, provide a practical guide and tool for deploying a Target - Evaluate - Improve loop versus baseline judges. The session will provide a clear workflow, a concrete demo, and a take-home solution to evaluate and improve your agent.

For whom: It's for those who want to be on top of the agentic game in 2026. We will cover how to define success targets, evaluate real agent runs at the trace level, and route each failure type to the right fix, so evaluation feeds directly into improvement. You will get a precise tool to evaluate and improve your agentic systems, along with:
- The latest landscape update on agent evaluation frameworks, what works, what doesn't, and major updates in 2025-2026
- Walkthrough of the leading optimization approaches and where they fall short
- Hands-on review of the TEI (Target, Evaluate, Improve) tool and the specific problem it solves: connecting evaluation output to automated, targeted fixes
- Live before/after demo on a real agent showing baseline output, per-dimension evaluation scores, applied improvements, and updated metrics.
- Practical guide on how to apply this to your agent.

About the Speaker: Orkhan Javadli, MIT'25, is a Co-founder at hAip (StartX), building multi-agent AI architectures. With over a decade of deploying architecture and production systems across telecom, e-commerce, and fintech, he focuses on building systems that turn agent evaluation into automated improvement at production scale.

​About Founder Spotlight:
This session is part of our Founder Spotlight series. It's a space for founders to showcase something cool they’re working on. Whether it’s a product demo, a behind-the-scenes look at their journey, or lessons learned, these sessions are meant to be informal, insightful, and founder-led. Join us to learn, ask questions, and support your fellow founders.