New article written on Collaborative RAG.

Click below to read on Substack.

Link to Post: https://habibkakakhel.substack.com/p/ai-trees-instead-of-ai-factories

Link to Post: https://habibkakakhel.substack.com/p/mlops-genaiops-and-agentops

Link: Take all my Data but Augment Me – by Habib Kakakhel

https://www.theverge.com/2025/1/13/24342683/mercedes-benz-mbux-virtual-assistant-google-automotive-ai-agent

Noting that Google has released ‘vertical agents’:

https://cloud.google.com/solutions/vertical-ai-agents#ai-agents-built-for-specific-use-cases

Two are released currently. One for Automotive and another for Food Ordering. They can be seen at the above link.

Microsofts Magentic One:

https://www.linkedin.com/pulse/what-makes-magnetic-one-unique-suparna–zy56c/

As part of my long journey in networking I’ve observed that networking has evolved to accomodate cloud and SaaS traffic. Be it Multicloud or be it SDWAN, networking has accomodated these new traffic patterns.

In this post I would like to highlight some essential items which I think to be the crux of networking.

1: IP Endpoints: Consider this to be the endpoint. Endpoint of what ? An endpoint which can be considered as a source or sink of traffic. An endpoint from which traffic is generated or where traffic is sinking into. Or it could be an endpoint where the journey of traffic translates into another journey.

2: Block of IP Endpoints: These are subnets or CIDR blocks. They are blocks or groups of IP endpoints which are used to communicate the existence of a group of IP endpoints in a certain locations. So for example x.x.x.x/24 will be in North America while y.y.y.y/24 will be a group of endpoints in Asia.

3: Routing Domain: This is an imaginary logical boundary behind which a group / block of IP endpoints are being routed. Consider a routing domain to be a domain which has a routing table. A routing table which lists the groups of IP endpoints inside the routing domain and where they further connect to at Layer 2.

This is it and I will stop at these 3 only. Why ?

This is because the classic networking in Service Providers or Enterprise WAN networks has evolved to accomodate SDWAN and Cloud so much that I think we need to stop at these 3. The (older) concepts of routing protocols are now considered solid parts of networking which are a given. A routing protocol is sprouted so as to exchange the routability of blocks of IP endpoints. The reachability information is shared ahead so that the IP endpoints and their blocks/groups are reachable from far away. They are a given now.

Why only these 3 ?

This is because I feel most of networking can be explained starting from these three concepts. For example concepts of end to end reachability, a route table, filtering at the edge of the routing domain, firewalling for IP endpoints and layer 4, etc etc can be built and explained around these three concepts.

These are the three concepts that are needed to explain the cloud networks in the form of Azure VNets, AWS VPCs and GCP networking.

I know there are other concepts to networking like Load Balancing, NAT etc etc. But I like to think that these three are the crux of networking. At least for Layer 3.

Thanks, Habib

Analyst firms like Gartner have given indications that soon there will be digital assistants, digital employees, digital co-workers and/or advanced virtual assistants in the workplace. All these terms signify a similar trend. A technological bot like tool as a full co-worker. I will use the term Advanced Virtual Assistant.

In this write-up I intend to connect two phenomenon and give an outline of what an advanced virtual assistant for IP Networks might look like. The first phenomenon is Hey Google, Apple Siri and Amazon Alexa. These are personal virtual assistants which listen and respond to inputs from people. The second phenomenon is the open space available for a virtual assistant helping run IP networks and technology systems.

The basic example of a virtual assistant is that of Siri, Hey Google and Alexa. These are devices which listen and respond like a human. They require computation and processing to listen and calculate what is being said. Thereafter they require computation and processing to formulate a response and send it through. The I/Os are a mic to listen, a speaker to speak and communications over the Internet.

The way I see it is that every large enterprise and large network which is run by multiple network engineers could have a trained virtual assistant. Some of the basic repeated tasks of a Network Engineer are:

• Run a ping or a traceroute on an IP
• Login to a switch, router or firewall and run some commands
• Look up an ISP circuits details from the ticketing system.
• Look up routes in a router for a subnet
• Look up empty IP address space in the IPAM. IP Address Management system.
• etc. etc. etc.
  These are repeated actions regularly performed again and again by all Network Engineers throughout the world.

The question is: How could an Advanced Virtual Assistant help in these repeated actions?

From here on imagination starts. Imagine being able to speak to a bot on your work laptop. Lets call the virtual assistant Netty. Imagine being able to say Hey Netty, run a traceroute for 12.2.2.2 or Hey Netty, please you login to the route isbpk56. Imagine thereafter getting Netty to do this on the router: Hey Netty, run the command ‘show ip route 2.2.2.2’, or something like Hey Netty, run ‘show inventory’.

So basically, you have a Hey Google, Siri or Alexa type assistant for your network. It has been fed the networks IP addresses, limited rights login details, router names and other details as required. Now you are talking to it and it is doing some work. This is my understanding of ‘digital co-worker’ and ‘advanced virtual assistant’. An assisting device which you can talk to and it is able to execute the actions.

I feel that instead of large virtual assistants covering big domains there will probably be industry specific and audience specific virtual assistants. Even probably they will be company specific, trained and modified per organisation.

Let’s keep imagining. Imagine having a search option as well. For example: Hey Netty, search the network for any firewall rules on 10.62.45.3. This would be very handy when looking for some firewall rules. Or Hey Netty, search the network for the mac address. abcd.efds…

Similarly imagine getting a complaint for a circuit and saying Hey Netty, please bring up the circuit TX459’s details on the ticketing system. Or Hey Netty, search the network for the interface descriptions having the circuit TX459 in them.

I think this has a great potential to improve efficiency and productivity. If Network Engineers get a Hey Netty for their networks, they would probably be able to handle a lot of work very easily.

As expected, this will require work to get a conversational bot which is a good bot listener and a good bot speaker. There will also be search and response algorithms, databases queries and responses. etc. etc. That will be the work done to make Hey Netty possible. But it is an engineering exercise which would give much productivity increase especially for large networks run by large teams.

Similarly, the same is the case for Storage and Servers technology domains or even Telecom networks. The Storage engineers will have their own regular instruction sets, the servers and systems their own and telecom network their own. They could all get bot like virtual assistants trained to carry out actions.

Imagine saying this and getting the work done:
Hey Techy, could you show the details of the VM sydau78
Hey Telly, could you bring up the RF stats for the site melau86
Hey Netty, could you ping to the ip 3.3.3.3

I feel the industry leaders should step in to bring this into open standardisation efforts. So, for example Storage, Networking, Systems, Telecom industry leaders should each formulate teams to make standard Hey Netty like query-response lists. Thereafter bots could be trained for Hey Netty v1 and Hey Netty v2 as the query-response details come in from the standardisation body.
It is time for Hey Techy and Hey Netty to hit the technology space. It is time for the post-pandemic knowledge workers to also get advanced virtual assistants and digital co-workers. I feel this is a good time for standardisation efforts for these as well.

Thank you,
Syed Habib Kakakhel

A question very important is what will happen in the world when bandwidths become very high. Initially there was GSM call and DSL, then came 3g and fibre and now there are higher bandwidth fibre and 4g. There is 3g/4g and Wifi based fibre broadband in most places now.

To track the effects of bandwidths we can take a back to basics approach. It is that humans have eyes and ears. So it boils down to sight and speech. Therefore we can say that very high bandwidths like 100 Mbps in telecom and 1 Gbps on Wifi broadband will result in better things we can see, we can hear and talk to.

I feel 3D VR, AR and 4k, 8k video covers sight. In terms of speech and sound I feel voice operated devices and devices speaking interactively will become possible. Like Alexa, Siri and Hey Google.

Humans are limited in terms of seeing and hearing. Perhaps touching and neurological signals can be included but for now the eyes and ears are the primary interfaces that humans use to interact with technology.

Lets now take it further. With greater bandwidth and greater processing comes greater life-like responses from computers. So we there could be voice assistants assisting humans with all sorts of productive capabilities. Imagine you have a secretary and you can ask them to do stuff. Imagine now that there is a computer that is processor and bandwidth supported making it equivalent to a secretary and even more efficient. If one wants to schedule a taxi ask the device secretary, if one wants to schedule a flight ask the device secretary, if one wants to buy grocery ask the device secretary, if one wants to order something ask the device secretary. These are being called Virtual Assistants (Gartner predicts that, by 2025, 50% of knowledge workers will use a virtual assistant on a daily basis, up from 2% in 2019). So everyone will have a mini secretary in their smartphones and people who want to pay more will perhaps have sophisticated virtual assistants.

That’s ears, voice and speech.

Lets discuss eyes and sight. With greater bandwidth and greater processing comes the ability to show 3D life-like views to people (if they wear a headset and enter that domain). So perhaps a 3D camera will take in a 3D view of a meeting room and given sufficient bandwidth it will transfer that 3D view to another place. Someone wearing a headset can see a 3D view of any place and will really feel like he is there. Voice can be part of the transfer so people could wear a headset and be seeing and hearing a 3D view of a far of place. Instead of a 2D video call it’ll be a 3D view call. This makes way for new avenues in human interaction. So imagine an MIT classroom having a 3D camera and anyone in the world attending the class in 3D view via a headset. It will feel very much like a real class attendance and will definitely help learning. I think it will make a much better learning experience than the 2D video one because of its feel and presence effect. Its easy to get distracted into something other than what you are seeing on your 2D screen but a 3D view would be immersive learning. This is being called immersive learning and experiential training because it is immersive in experience and is not just sight but is an experience.

Enter The Matrix. You wear a headset and someone else wears a headset and they could be in a 3D virtual world together. Lets not make them fight here on the blog but two people being together in 3D view is a very real possibility given sufficient bandwidth and processing.

So now lets break this down. Consider a 3D live transmission and people attending in 3D view that live transmission. As in an MIT Class which people across the world attending in 3D view LIVE. Now separately consider non-live transmission. A virtual 3D world which 2 or more people are present in together which is a recorded experience. For example a recorded Safari trip which is not live but people jointly attend it. Now separately consider a computer generated 3D world in which there is a virtual assistant like a device secretary present as well. Someone wearing a headset is taken to a computer generated 3D virtual world where they are with a virtual assistant, are talking to them and learning from them etc. Now further consider 2 or more people being in a computer generated 3D world together and there being a virtual assistant present in that world as well which they can speak to. Interesting possibilities.

These are real possibilities given sufficient bandwidth and processing. I feel 5G and Wifi 6 causing bandwidth increase can result in these. Simultaneously edge computing and cloud computing causing increased processing can bring these things to life. These seem like real possibilities and things are progressing in this direction.

I came across a job ad for a Network Consulting Engineer. Its worth a blog because it shows whats happening in the market. The Ad states:

Alpha (renamed) is looking for an experienced Network Consulting Engineer to join during this time of growth.

Key responsibilities are set out below:

• Consult with clients to design and implement network security technologies for example next-gen firewall, remote access, network access control, SaaS and public cloud networking and security services.
• Work with a keen eye for detail within a network and security context across public, hybrid and private cloud environments
• Proven network engineering project background with knowledge and skills in analysis, design and implementation for some or all of the following areas: LAN, SD-WAN, WLAN, Cloud and Network Security Technologies
• Technical oversight for complex projects incorporating multiple technology streams.

Comprehensive subject matter expertise with at least 3 of the following technologies:

• Routing and Switching solutions
• Firewall and Network Security technology
• Wireless LAN infrastructure and RF Design
• SD-WAN and SASE Solutions
• Network Access Control and Authentication systems
• Cloud Networking technologies – AWS and Azure

The additions to the networking landscape is obvious. The old is still present. Routing and Switching for covering IP routes, LAN for L2 switching and Firewalls. These are stable technologies covering layers 2,3,4 communications and network security. They include MAC address tables, vlans, IP addressing, subnets, route tables, LAN, WAN, SP VPNs, Core/Agg Data Center, TCP/UDP port filters and access-lists amongst other things. This is basic networking.

Further after this is Wireless LAN infrastructure and RF Design. This covers Wifi Access Points and WLCs. So for example an enterprise could be upgrading there wireless coverage. To cover an area like an office RF design planning is required which includes RF signal strength considerations for AP placement in floor plans.

Further after this is SD-WAN and SASE. These are the new blokes in town due to changes in the landscape of networking. Two things cause SD-WAN:

• Private MPLS WAN links being expensive

• Public internet connections having become faster

Due to this WAN is shifting from MPLS VPN SP links to internet back-hauled links. This requires:

• new branch edge devices
• new branch edge design to include internet back-haul
• new HQ design to include internet and SD-WAN HQ networking

That’s SD-WAN.

SASE is a Gartner term caused by new traffic sources and new traffic sinks. The enterprise traffic patterns have changed a bit. Traditionally all enterprise traffic would come to the HQ via a WAN and then it would go to the internet via a firewall from there. Enterprise applications would also be hosted in the HQ. Now increasingly cloud based applications are directly being accessed from the branch via internet connections and connections to the public cloud. This means new traffic patterns are present where the branch edge is talking to cloud applications directly without the HQ in between. Enterprise applications are now also in the cloud instead of an HQ PoP and the edge is the branch. Gartner saw this new traffic pattern and suggested that there needs to be secure access at the service edge in the branch (SASE).

For any enterprise which wants to move towards an internet based WAN network and has also moved its applications to be cloud-based then their whole traffic patterns have shifted away from the HQ PoP. This means these enterprises will go for SD-WAN and SASE solutions to use internet-backhaul and to secure the service edge. This is all new but makes sense and with sufficient push they can be implemented. Most work required for these will be:

• new branch designs for SD-WAN internet back-hauled branches and
• new HQ POP design to include internet-based SD-WAN HQ.

There are SD-WAN vendor providers which offer these MPLS WAN replacement solutions to either augment the MPLS WAN or replace it.

Most of the SASE work will be solution evaluation and vendor assisted implementation via GUIs as I see it.

Next on the list is Network Access Control and Authentication Systems. As I see it this is 802.1x, Radius Authentication and could also include Active Directory based Authentication/IAM. From a networking perspective 802.1x will need to be enabled on the LAN ports. Perhaps Radius based authentication will need to be integrated in systems and perhaps firewall rules will be required for Microsoft AD access.

The final item on the list is Cloud Networking which includes AWS and Azure connectivity. From my experience in hybrid and multicloud deployments much of the work is integrating new links into existing networks and configuring routes establishing IP connectivity between new endpoints. So some IP subnets and IP endpoints are on-prem and some are in the cloud. Routes are required and links are required. It still requires Layer 1 networking with new links and SFPs for say Megaport or Direct Connect. It still requires a semblance of Layer 2 where across the megaport or direct connect layer 2 is reachable. On top of these routing reachability is established with routes to be added in the relevant locations. So for example an on-prem firewall could be an L3 routing SVI location and routes pointing towards the cloud would be added there. Similarly routes would be added to routers and propagated so as to provide the cloud subnets reachability inside the on-prem network. On the other side in the public cloud side the VPC/Vnet would have routes pointing this way. There would be configuration items establishing the link between the relevant AWS direct connect or Azure expressroute to the VPC/Vnet. If its multicloud there could be multiple AS’s involved and BGP would be used to establish route exchange.

Most work on AWS/Azure side is GUI based unless Infrastructure as Code is used for large deployments. If infrastructure as code is used then Ansible, Terraform, Git and Bitbucket will be used instead of the GUIs to configure the routes and the the firewalls in the public clouds. Infrastructure as code has two main sections, one being version control of the config code and the other being the deployment of updated code to the public cloud.

The other non technical requirement in the job ad is:

This is standard MSP or enterprise project work where design documents, implementation plans, standard configs and wiki articles are part of the project.

Together all this makes a Network Consulting Engineer.

First some definitions and news relating to the subject:

Process Mining:

Process mining is a family of techniques relating the fields of data science and process management to support the analysis of operational processes based on event logs. The goal of process mining is to turn event data into insights and actions.

Digital Twin:

A digital twin is a virtual representation that serves as the real-time digital counterpart of a organization or process.

News:

https://newsroom.ibm.com/2021-04-15-IBM-to-Acquire-myInvenio-to-help-Organizations-Use-AI-powered-Automation-to-Better-Streamline-Business-Processes

Background:

myInvenio is a disruptive solution of Process Mining and Digital Twin of an Organization to automatically analyze business processes, monitor and predict the future trends of processes.

Details:

“Addition of myInvenio provides customers with a comprehensive AI-powered automation portfolio for business automation – including process mining, robotic process automation (RPA), document processing, workflow, and decisions”

Notes:

Given the above details regarding Process Mining within an organization and creating a Digital Twin of an organization and then moving towards greater efficiency it is obvious that Network Automation is one piece of the puzzle. Efficiency of the whole organization depends on analyzing the processes and then putting the right kind of automation in place. Reading up on another article I saw that one method of process mining is to conduct interviews across organization to analyze what the staff do and to analyze the human part and the machine doable part. RPAs, Network Automation, Systems automation are one part of the puzzle. Gartner is calling this Hyperautomation.

Details:

Hyperautomation is the application of advanced technologies like RPA, …. and Process Mining to augment workers and automate processes in ways that are significantly more impactful than traditional automation capabilities. Hyperautomation is the combination of automation tools to deliver work.

Gartner’s report notes that this trend was kicked off with robotic process automation (RPA). The report notes that RPA alone is not hyperautomation. Hyperautomation requires a combination of tools to help support replicating pieces of work where a human is involved in a task.