Talk to Sumanth about
How Sumanth thinks
Evaluation separation
Separates semantic judgment from scoring logic, making large‑scale agent evaluations reproducible.
For you: You can get consistent scores by defining decisions up front.
Reliability over speed
Prioritizes fetch reliability before speed to ensure consistent page retrieval across domains.
For you: You should favor reliable fetching when building agents that need consistency.
End‑to‑end voice booking
Builds open‑source voice AI agents that manage the full appointment booking workflow via telephony.
For you: You can deploy similar agents to automate phone‑based scheduling.
Ask this first
- How does JevEval ensure deterministic scoring?
- What challenges did you face building a voice AI booking agent?
- Why prioritize fetch reliability over speed in agent design?
Sumanth's brain
Recently updated
In mid‑October 2026 I introduced **JevEval**, a deterministic evaluation framework that uses Jev as the judge instead of a generative LLM…
Decision: prioritize fetch reliability over speed for agents that need consistent page retrieval across diverse domains.
One MCP connection for actions across multiple apps