Eduard (Edward) Dulharu

speaker

420 appearances 1 recordings 1 series first heard Aug 2026 last heard 7 Aug

Eduard (Edward) Dulharu’s voice in public audio — every appearance, attributed to the second.

Trend

recordings per month · last 12 months

Recordings per month over the last 12 months — 1 in all, peaking in Aug 2026 with 1.

Appearances

newest first · ▶ plays the moment

Heavy Networking · HN837: Agentic AI to Reduce MTTR · 7 Aug 2026 podcast

I think agents should be confined within a very precise domain of investigation, should not have access to everything.

Uh if an for example, if an agent uh a routing agent is in charge with fixing uh routing problems, he should not have access to the firewalls or to to the firewall rules.

It's simply uh this option does not exist within the YAML file, which is the source of truth.

The option to to land on a firewall does not exist for that agent.

Yeah, I think both uh angles of uh analysis are are valid because um well the idea is to have um um controlled and deterministic uh approach.

So the agents should act as um force multiplier, not as uh unpredictable

tools which cannot predict an engineer should be able to predict what will will be the um what will be the outcome of an agent.

I mean the engineer which knows that YAML file, which is the sum of all the knowledge.

should be able to predict what uh what what are the actions of an agent, which means that by design the uh um

By design, this workshop should be deterministic, should and should be guarded.

The model should not be allowed to uh fabricate anything, should take in uh uh uh quality data and should provide uh an should reason against and should provide an answer.

So the agent should have limited uh

Limited capabilities.

No, it's not the case.

So um

The only um uh so the the idea is that an an AI system should not replace things which work today, like uh

uh deterministic workflows, coding, etc., should add the power of uh a large language model or a small language model fine-tuned to capture unknown things, things which cannot be um which cannot be uh captured within uh classical programming.

So uh this is basically where um um with models uh

Excel is to catch corner cases, but for everything which can be done deterministically then um we do it uh

In the old school fashion, so uh and another point uh w within this context is that um what classical programming misses is that um

Showing 381–400 of 420 · page 20 of 21