---
session_id: "1373460"
title: "Heal Your Agents, Don't Just Grade Them"
phase: live
last_checked: 2026-11-12T13:24:21-08:00
render: 20261112-132421-c4d6ee
---

# Links: Heal Your Agents, Don't Just Grade Them

## Speakers

**Khasimul Asif Mohammad** ("Asif"), Principal Product Manager, AI Observability &
Resilience, Splunk

**Sudhir Tonse**, Director of Engineering, Core Platform & Research for AI
Observability and Resilience, Splunk

Neither profile could be confirmed. Asif is a co-author of the Signals paper
below.

## Company

**[Splunk Agent Observability](https://agent-observability-docs.splunk.com/)**:
"Splunk Agent Observability is the leading observability, evaluation, and
guardrail platform for GenAI and agentic applications." (from the docs; formerly
Galileo, per the roster)

## Resources

- **Paper** · [Signals: Continuous Diagnostics of AI Agents](https://arxiv.org/abs/2611.04507)
  (arXiv, 2026-11-03, co-authored by Asif Mohammad). The method behind the talk:
  a taxonomy of 45 problem types in eight families, findings grouped into
  Signals with remediations, and 91.43% F1 against reference issues.
- **Paper** · [ChainPoll: A high efficacy method for LLM hallucination detection](https://arxiv.org/abs/2310.18344)
  (arXiv, 2023-10-22). Galileo's paper, where the talk's journey from
  LLM-as-judge to Signals starts (transcribed "ChainPole").
- **Product** · [Signals docs](https://agent-observability-docs.splunk.com/concepts/signals/signals).
  Sweeps production traces, groups related findings, and ranks the failure
  patterns that evaluators miss.

## Gaps

- LinkedIn (`linkedin.com/in/khasimulasif`, from the roster) answered 999 with no
  page title naming him, so it is not corroborated and is left out.
- No profile for Sudhir Tonse could be verified. The roster lists none.
- splunk.com answered 403, so the company links the Agent Observability docs.
- The small evaluation model contrasted with an OpenAI release (transcribed
  "Guna") was not identified.
- The published efficiency and quality datasets the talk mentions were not found.
- Galileo's "Context Engineering at Scale: How We Built Galileo Signals" (listed
  here until this recheck) stopped answering: galileo.ai now redirects every page
  to Splunk's Agent Observability product page, and no Splunk copy was found.
- No official recording or slides found yet. The dossier rechecks every 6 hours
  during the event.