Karthik Mahalingam writing at LatentMesh

How AI systems fail.
How to evaluate them.
How to govern them in production.

Most AI governance advice skips the engineering. These essays don't.

Read latest essays Start with the series
19+ years building distributed systems at scale
Engineering leader at Microsoft
Writing on AI systems, evals, safety, and governance

Series: Reliable Agent Systems · Ten essays

Start here

New to LatentMesh? Start with the core essay series on how agent systems fail, how to evaluate them in the real world, and how governance becomes operational.

Agent Failures Are Distributed Systems Failures

Why agent failures are usually system failures, not just model failures.

The Eval Gap

Why staging success does not predict production reliability.

Controls Are Not Guardrails

Why runtime safety layers are not enough without evidence, ownership, and testing.
Read the series →

Latest essays

Recent essays on agent reliability, evaluation, safety, and governance.

The Regulatory Mapping Table An interactive reference that turns EU AI Act high-risk obligations into operating controls, verification methods, evidence artifacts, owners, and review cadence. Filter by role, article, cluster, or cadence to map obligations into your operating responsibilities. What Your Agent Logged vs. What the Auditor Needed The trace says what happened. The auditor asks why, under what authority, and what changed. Most agent deployments log enough to debug a success but not enough to investigate a failure. Drift Detection Patterns for Production Agents Your agent is still answering. That does not mean it is still behaving the same way. Five drift classes, three detection layers, and the patterns that catch regression before your customers do. From Obligation to Evidence in 90 Minutes Pick one requirement. Map it to a control. Write the eval. Generate the artifact. Assign the owner. A hands-on walkthrough of the full compliance loop using EU AI Act Article 14. Building an Eval Harness That Survives Production Most eval harnesses die the same way. Five structural decisions separate the ones that survive production from the ones that quietly rot. The Incident Response Gap in AI Systems You built the controls. You still cannot contain the failure. Most organizations have started building AI controls. Far fewer have built AI incident response.
View all essays →

What I write about

Agent failure as system failure

How agent systems break, drift, and misbehave once they leave the demo.

Evaluation that catches real failures

Practical evals for regression, reliability, and operational trust.

Safety as systems engineering

Guardrails, control surfaces, observability, containment, and human review.

Compliance that becomes architecture

How legal and policy obligations turn into workflows, controls, and evidence.

About

I write about AI systems from the perspective of someone who has spent nearly two decades building large-scale distributed systems. My focus is on how agents behave in production, how evaluations catch real failures, and how safety, privacy, and compliance become operational reality.

More about me →