RAG‑Powered Troubleshooting for NOC Engineers
Key Points
- Rebooting is often a quick fix, but skilled engineers need to identify root causes and apply precise solutions.
- Retrieval‑Augmented Generation (RAG) combines vector similarity search with large language models to let NOC engineers quickly pull relevant documentation, tickets, and FAQs.
- The workflow involves chunking source texts, embedding them into a vector database, retrieving the most pertinent pieces for a query, and having an LLM generate a concise answer.
- LLMs can also auto‑summarize past incidents and similar tickets, streamlining issue analysis and even assisting in creating new trouble tickets.
Sections
- Using RAG for NOC Troubleshooting - The speakers explain how Retrieval‑Augmented Generation with large language models can help network operations engineers sift through documentation, tickets, and FAQs to pinpoint root‑causes instead of just rebooting.
- AI-Driven Trouble Ticket Automation - Discussion on using generative AI to auto‑populate ticket fields, classify priority and severity, and assign tickets to the appropriate team, streamlining NOC engineers' workflow.
Full Transcript
# RAG‑Powered Troubleshooting for NOC Engineers **Source:** [https://www.youtube.com/watch?v=1r4CIjNPU0o](https://www.youtube.com/watch?v=1r4CIjNPU0o) **Duration:** 00:05:53 ## Summary - Rebooting is often a quick fix, but skilled engineers need to identify root causes and apply precise solutions. - Retrieval‑Augmented Generation (RAG) combines vector similarity search with large language models to let NOC engineers quickly pull relevant documentation, tickets, and FAQs. - The workflow involves chunking source texts, embedding them into a vector database, retrieving the most pertinent pieces for a query, and having an LLM generate a concise answer. - LLMs can also auto‑summarize past incidents and similar tickets, streamlining issue analysis and even assisting in creating new trouble tickets. ## Sections - [00:00:00](https://www.youtube.com/watch?v=1r4CIjNPU0o&t=0s) **Using RAG for NOC Troubleshooting** - The speakers explain how Retrieval‑Augmented Generation with large language models can help network operations engineers sift through documentation, tickets, and FAQs to pinpoint root‑causes instead of just rebooting. - [00:03:04](https://www.youtube.com/watch?v=1r4CIjNPU0o&t=184s) **AI-Driven Trouble Ticket Automation** - Discussion on using generative AI to auto‑populate ticket fields, classify priority and severity, and assign tickets to the appropriate team, streamlining NOC engineers' workflow. ## Full Transcript
Sai, have you ever run into a technical issue
with an electronic device?
And the only solution you could think of
was to actually reboot it?
Yes Sharath, that's my first thought,
to reboot the system.
And, it usually magically fixes the issue.
But a skilled engineer might have to actually find out
what is the main root cause,
and then find a particular fix for that issue.
How would one go about doing that, because
there could be a lot of reasons why
a particular system or device or network is failing.
How can we do that?
We can use large language models.
You know, the there's a lot of buzz about generative AI
to actually do that.
So that is a technique known as RAG,
or Retrieval Augmented Generation,
that can help a NOC engineer or any other skilled engineer
quickly find the right information
to help solve a particular issue.
I've heard about generative AI and LLMs,
but in this scenario a NOC engineer will have a lot of documentation -
a lot of product documents, troubleshooting tickets,
troubleshooting information to go through.
How can a NOC engineer go through all of them
and come up with a solution?
Yeah, so let me explain this RAG framework a little bit for you.
So, as you mentioned, there are a lot of different data sources.
We have product documentation.
We have previous incidents that might be similar.
We also have FAQ documentation, trouble ticket information.
So a lot of different diverse set of data sets
which a NOC engineer might have to go through.
But we can actually chunk all of this data,
then we can run it through an embedding model
and convert this text data into numerical values
and store those in a vector database.
Now, when a NOC engineer has a question,
he or she can ask that question to the vector database.
It can retrieve the most relevant information
and pass that information over to a large language model.
The large language model is able to then generate an answer
and send that back to the NOC engineer,
who can then use that to solve the issue.
Okay, so this is how the application will look like.
The NOC engineer basically asking a question.
And we use all of the data that we have
in terms of incident related data, trouble tickets,
and we get the relevant information, relevant incidents back
and it's displayed back to the NOC engineer.
So, how else can we use generative AI and LLMs
in NOC scenarios?
Sure, so another thing we can use is to generate summaries
or summarization of previous incidents
and similar tickets that were in the past.
We can create a summary of those using a large language model
and present that information to a NOC engineer.
Okay, so the NOC engineer, all the NOC engineer needs to do is
type in a question and get all of the relevant information,
and also get the summary basically to look at
what could have happened in that particular instance.
That is good.
But in certain instances where
the NOC engineer needs to actually create a trouble ticket
to be able to solve a particular issue,
can we have generative AI help in that as well?
So we can also do things like
generation of particular fields in a trouble ticket.
So let's say we have a trouble ticket with a whole lot of fields
where there are some text fields, some other fields that
a large language model can automatically pre-populate
for a NOC engineer to quickly review and submit information.
That is great.
So all the NOC engineer should look at is the
created auto-generated trouble ticket
to look at all the fields that were pre-populated
and just review them.
But also, how do we define which team this
trouble ticket gets assigned to and the kind of severity
and that kind of complexity for that trouble ticket?
Can we have generative AI help in that aspect?
Absolutely, so large language models can also do classification.
So, based on previous and similar incidents,
a large language model can decide which is the best team
to solve that particular issue.
It can also classify priority
as well severity based on similar previous incidents.
Okay, that is great.
So using all of this information,
the NOC engineer will be able to use this information,
create the trouble ticket and submit them
so that it gets assigned to the right team.
But let's say in this in this scenario,
the engineer or the developer who's picking up this ticket
needs some help with creating a fix for this issue?
Absolutely, so we can also do code generation,
to actually generate the code that's required to fix the issue.
Code generation, that that looks interesting.
Can you explain more about code generation and how can we
use generative AI in that?
Sure, so there are a lot of large language models
that's trained specifically on code.
So it can actually generate code
based on, you know, troubleshooting steps.
Or it can also assist a developer who's writing the code
to do code completion and other things
so that the developer can quickly write the code that's required
to solve the particular issue.
Okay, that is great.
So the developer needs to look at the code, review it,
maybe import it into his environment,
into his or her environment and just be able to execute that
and be able to fix this issue.
Absolutely.
That's amazing.
Yeah, so next time you have an issue,
don't just think of rebooting it.
You can actually use generative AI
to do a whole lot of these things,
and then actually find the root cause and solve the issue.
Absolutely, using all of this I can definitely do that.
If you liked this video and want to see more like it,
please like and subscribe.
If you have any questions or want to share your thoughts about this topic, please leave a comment below.