Home Blog

AI Community Cisco Buyer Voices – Colt, STC and Swisscom


I sat down with senior leaders from Colt, stc Group, and Swisscom to listen to these modern service suppliers’ methods for the supply of digital experiences in an AI-powered world. We talked about how they’re simplifying their networks to satisfy rising buyer calls for and the way they’re embracing assurance, automation, and autonomous networking to rework their operations, make sure the resiliency of their providers, and drive new income streams.

Q: Jason Teller—Yasser, congratulations on the success of the 2024 Esports World Cup. Share with us the method and applied sciences stc applied to make sure a seamless expertise for gamers and viewers.

A: Yasser Alkholaif, Transport Operations Director, stc GroupGiven the dimensions of the Esports World Cup, there was lots on the road. To make sure a seamless expertise for gamers and viewers we would have liked real-time visibility throughout the community so we might proactively deal with points earlier than players even seen.

We arrange a specialised operations heart geared up with superior monitoring instruments built-in with Cisco Supplier Connectivity Assurance (previously Accedian Skylight) and ThousandEyes. We constructed customized dashboards, configured real-time alerts, and established automated ticketing for important metrics—latency, packet loss, and hyperlink utilization.

This gave us AI-powered subject detection and proactive assurance throughout our infrastructure, worldwide carriers, and cloud supplier networks. Ultimately, we delivered a top-tier gaming expertise for all 2024 Esports World Cup individuals, with no main disruptions all through the occasion.

Q: Teller—That’s an incredible achievement, particularly given the large quantity of strain concerned with international sports activities occasions. Martin, I’m eager to listen to your ideas on the significance of high quality of service for Swisscom. How has it impacted the best way you handle and function your community?

A: Martin Gysi, System Architect for IP Networks and Community Automation, SwisscomWe’re important infrastructure for Switzerland. Our community is the inspiration for the nation’s emergency providers. Tens of hundreds of companies depend upon us to have the ability to function successfully and serve their prospects, and our cellular and residential prospects anticipate flawless digital experiences.

Beforehand, our infrastructure was extremely complicated, with 4 purpose-built networks for cellular, enterprise providers, residential providers, and knowledge facilities—all operated by completely different groups. To simplify our infrastructure, we’re consolidating these 4 networks and groups into one, which can drive operational effectivity and resiliency. Section routing and having a single IP companion—Cisco—play a key function in that.

We now method the community as if it was a chunk of software program, and our ambition is to automate every little thing. For instance, we’ve automated all our take a look at instances and now configure the community by means of Cisco Crosswork Community Automation. We’re additionally adopting community as a service (NaaS), with a clearly outlined set of providers that may be ordered by means of APIs. We’re at the beginning of an extended journey however we all know it’s going to repay—for us and our prospects.

 

 

Q: Teller—Simplicity, visibility, and management over providers are all essential for resiliency. Bart, you may have introduced all of it collectively in Colt’s NaaS provide too. Inform us about your NaaS journey so far.

A: Bart Janssens, Principal Packet Architect, Colt—The NaaS innovation journey on our IP community began in 2021. As the largest pan-European aggregator, we would have liked simplicity, capability, and scalability, whereas managing CapEx and OpEx and making certain the facility effectivity of our infrastructure. This led us to Cisco Routed Optical Networking. We have been the primary service supplier to undertake it on lengthy distance core spans.

We then deployed a software-defined networking (SDN) stack to regulate our {hardware} and differentiate our providers, and unified APIs to our OSS/BSS to orchestrate them. After deploying Cisco Crosswork Community Automation and Supplier Connectivity Assurance to gather knowledge throughout our community, we created our personal NaaS enhancements that we name Intent-Primarily based and Sustainable Networking.

We all know that community demand will develop due to AI. Prospects will need extra capability and stringent SLAs for AI-powered chatbot brokers, collaboration instruments, hybrid workloads, on-premises and multicloud, storage administration, and studying fashions.

Our Intent-Primarily based Community offers enterprise prospects unprecedented visibility and management in actual time from wherever. They will see and management an supposed path hop-by-hop, by latency, and by its carbon footprint relying on the place routers are situated and the facility they use. Prospects can resolve what sort of service greatest meets their wants primarily based on what’s essential for his or her enterprise—be it SLAs, value, or sustainability.

 

 

Q: Teller—Automation and assurance capabilities in Colt’s NaaS have reworked the way you serve your prospects. Yasser, stc has embraced automation too, and is now transferring towards a extra autonomous community. Inform us in regards to the outcomes you search.

A: AlkholaifWe have now two objectives: decreasing operational prices and enhancing buyer expertise. By AI-driven insights and proactive automation, we glance to reduce guide interventions and streamline service assurance.

This is not going to solely scale back overhead but in addition enhance reliability, resulting in fewer disruptions and sooner subject decision. We’re already seeing outcomes from pilot initiatives, pushing us nearer to a self-optimizing and self-healing community that delivers a sturdy, responsive buyer expertise, whereas holding operational prices in test.

 

 

Q: Teller—TM Discussion board measures the maturity of autonomous networking with ranges from zero (guide operation) to 5 (full autonomy). TM Discussion board estimates 84% of service suppliers are at degree 1 (assisted operations and upkeep) or degree 2 (partial autonomous community), whereas 61% purpose to attain degree 3 (conditional autonomous community) or above by 2028. Martin, what’s the one studying you’d share with different service suppliers?   

A: Gysi—It’s worthwhile to have perseverance and a transparent objective. In case you begin with out realizing precisely what your purpose is, who you’re designing a brand new system for, and all of the associated dependencies, you’ll probably find yourself with a system that provides complexity somewhat than eliminates it. In case you method automation step-by-step, you’ll get there ultimately.

Actually take into consideration what you need to obtain from a enterprise and technical perspective. Ensure you clearly perceive your present programs and processes, and the way folks work—then decide one of the best ways to automate all of it.

Q: Teller—Service suppliers’ networks are important to connecting AI purposes. With out these networks, there isn’t a AI. Bart, what recommendation would you give to different service suppliers to make sure their networks are prepared for AI? 

A: JanssensThe earlier that service suppliers prepare for AI, the higher. You want the appropriate constructing blocks to your community and the appropriate software program stack to orchestrate and refresh your community. You want knowledge and analytics to make it clever and the appropriate parts for transporting knowledge at scale.

Given the info visitors progress pushed by AI, it’s good to handle the facility effectivity of your community—due to OpEx and since your prospects have Scope 3 sustainability objectives to satisfy. Steady NaaS innovation stays essential.

Prospects are in every single place, however none of us can join the entire world singlehandedly, so we have to work collectively. The extra we work collectively and expose APIs, the higher positioned we’re to serve our prospects and earn a living from AI too.

The tempo of AI innovation is nothing like we’ve seen earlier than. All of us want to begin redesigning our networks now to make sure that we’re not left exterior of the rising AI ecosystem, however a key a part of it.

 

how AI is shaping the evolution of IP networks.

Share:

Palletizing frosting buckets with Robotiq + Alba

0


When your product is frosting and your purpose is to maneuver it—quick, clear, and with out breaking a sweat—you don’t need to depart something to likelihood. That’s precisely why one bakery producer teamed up with Robotiq and Alba to automate their bucket palletizing course of. The mission? Stack a wide range of frosting bucket sizes onto CHEP and GMA pallets with out guide labor or downtime.

Spoiler alert: mission completed.

Be taught Greenhouse Gasoline Accounting for IT with Cisco Networking Academy


Greenhouse fuel (GHG) accounting is changing into important throughout each business. Nevertheless, IT has typically been unnoticed of the dialog. That’s why local weather schooling chief One Level 5 Academy has partnered with Cisco Networking Academy to create the “Introduction to Greenhouse Gasoline Accounting for IT” course. This free course helps IT professionals purchase the talents to measure, report, and cut back emissions related to knowledge facilities, cloud infrastructure, and computing {hardware}—guaranteeing IT groups are geared up to assist their organizations’ sustainability objectives.

What’s within the “Introduction to Greenhouse Gasoline Accounting for IT” course?

One Level 5 Academy makes a speciality of serving to learners break into inexperienced careers, whereas Cisco Networking Academy focuses on getting ready learners for the IT area, making them a robust mixture. This entry-level course attracts from OnePointFive Academy’s complete eight-week sustainability expertise accelerator program and takes simply two to 3 hours whole.

It’s accessible as both self-paced or instructor-led and can be utilized as a standalone workshop or as supplemental studying, and contains:

  • 1 hour of video classes
  • 1.5 hours of interactive lab work
  • Teacher-led coaching accessible
  • Earn a digital Credly badge, shareable on LinkedIn

The course teaches you:

  • The fundamentals of GHG accounting
  • What Scope 1, 2, and three emissions are, and find out how to distinguish between them
  • The right way to conduct fundamental greenhouse fuel calculations
  • IT-specific emissions discount methods
  • Sustainability reporting and compliance necessities
  • An IT-specific case examine on an IT {hardware} producer

Who’s Introduction to Greenhouse Gasoline Accounting for IT for?

This brief course is good for:
• IT professionals seeking to construct inexperienced expertise
• Sustainability professionals attempting to know tech emissions
• Faculty and vocational college students in IT, enterprise, or environmental science
• Profession changers transitioning into ESG and climate-related roles

Do I must know something about GHG accounting?

No prior data of GHG accounting is required—simply curiosity and curiosity within the intersection of IT and sustainability.

Which jobs are GHG accounting-related?

This course aligns with a rising set of sustainability-related job roles, together with:
• ESG Information Analyst
• IT Operations Supervisor with a sustainability lens
• Know-how Advisor (local weather or ESG focus)
• Information Middle Sustainability Specialist

Is “Introduction to Greenhouse Gasoline Accounting for IT” tied to a Cisco certification?

Whereas Introduction to Greenhouse Gasoline Accounting for IT isn’t tied to a proper Cisco certification, it does offer you a validated Credly badge to show to employers you’ve discovered invaluable data and lays a powerful basis for additional examine.

This course additionally serves as an amazing launchpad for Networking Academy’s broader IT curriculum. Programs like Networking Fundamentals and Cybersecurity Necessities are pure subsequent steps when you earn your Credly badge.

Begin studying totally free at the moment

You will get began with the self-paced model of Introduction to Greenhouse Gasoline Accounting for IT at the moment or find an in-person, instructor-led course close to you.

What do you concentrate on IT’s function in sustainability? Do you see a niche in GHG Accounting for IT schooling? Tell us your ideas within the feedback.

 

A Extra Sustainable Future for Africa: Cisco’s Position in Bridging the Digital Divide

 

 

 

 

 

Join Cisco U. | Be a part of the  Cisco Studying Community at the moment totally free.

Comply with Cisco Studying & Certifications

X | Threads | Fb | LinkedIn | Instagram | YouTube

Use  #CiscoU and #CiscoCert to hitch the dialog.

 

 

Share:



Overland AI demonstrates full-stack floor autonomy for uncrewed breaching

0


Overland AI demonstrates full-stack floor autonomy for uncrewed breaching

Overland AI built-in its OverDrive software program stack and SPARK {hardware} into the Normal Dynamics Small Multipurpose Gear Transport. | Supply: Overland AI

Overland AI yesterday mentioned it has efficiently demonstrated its autonomous floor automobile built-in with its uncrewed aerial system, or UAS-capable payloads as a part of a joint breaching experimentation effort with the Sandhills Challenge and the twentieth Engineer Brigade of the XVIII Airborne Corps.

The Seattle-based firm accomplished the demonstration on the U.S. Military’s premier expertise innovation occasion, Challenge Convergence Capstone 5 (PC-C5).

Based in 2022, Overland AI mentioned it’s powering floor operations for contemporary protection. The corporate has performed greater than a decade of superior analysis in robotics and machine studying. That and its field-test ahead ethos enable it to ship superior autonomy for unit commanders.

Overland AI mentioned developed OverWatch, its intuitive C2 interface, to supply commanders with the exact coordination of autonomous floor techniques wanted for advanced missions to succeed. The corporate claimed that it has achieved the end-to-end integration of floor autonomy, from operator to impact. It mentioned it’s placing this functionality into the fingers of tactical operators at this time.

SMET autos achieve autonomy

Hazardous missions in austere and electronically denied environments require dependable and resilient expertise, famous Overland AI. It mentioned its SPARK autonomy upfit and OverDrive stack allow floor autos to navigate off-road with out GPS or direct operator management.

The corporate constructed its totally autonomous tactical automobile, ULTRA, in-house by integrating SPARK and OverDrive right into a modular and adaptable platform that’s presently in manufacturing. The U.S. navy is testing a number of ULTRA autos, which might carry 1,000 lb. (453.5 kg), transfer as much as 35 mph (56.3 kph), and cruise for 100 mi. (160.9 km).

Working underneath the course of the Sandhills Challenge, Overland AI built-in its full stack of autonomy {hardware} and software program into Normal Dynamics Land Methods’ Small Multipurpose Gear Transport (SMET) automobile. The expertise has additionally been built-in with the Textron Ripsaw and Polaris RZR.

The SMET was initially managed by way of a tethered hand distant. Now, Overland AI’s OverDrive software program stack, SPARK {hardware} infrastructure, and OverWatch tactical interface allow autonomous operation.

Overland AI reveals help of Sandhill Challenge targets

U.S. Military troopers from the twenty seventh Engineer Battalion, twentieth Engineer Brigade, skilled instantly on Overland’s system throughout PC-C5. The troopers realized to independently function, troubleshoot, and keep the totally autonomous platform.

By the top of the train, they executed breaching missions, efficiently beginning up the Autonomous Floor System, conducting missions, and shutting it down independently.

The demonstration included two payload integrations. The primary was a trailer geared up with a mine-clearing UAS designed for breaching a lane. Within the second, a Stratin Engineering-developed drone launcher mounted on the SMET to deploy small attritable drones.

“Troopers independently and efficiently operated our floor autonomy for 2 consecutive weeks throughout Challenge Convergence Capstone 5,” said Byron Boots, co-founder and CEO of Overland AI. “Breaching workouts proceed to display the maturity of {our capability} and the tactical advantage of eradicating people from a number of the most harmful missions.”

Overland AI instantly supported the Sandhills Challenge’s goal, corresponding to minefields, wire, and ditches, whereas enabling distributed operations throughout a breach. The twentieth Engineer Brigade now has the upgraded SMET automobile for continued testing, and Overland AI will present discipline help and a software program license underneath its contract.

PC-C5 is a key a part of the Military’s persistent analysis and testing of next-generation warfighting capabilities, together with cross-domain operations within the INDOPACOM (Indo-Pacific) theater. The Overland AI floor workforce a the train included specialists throughout {hardware} integration, autonomy software program, infrastructure, discipline operations, product, and program administration.

Overland AI mentioned it continues to advance the state of floor autonomy for protection, having beforehand secured an $18.6 million contract with the U.S. Military and Protection Innovation Unit (DIU) to develop autonomy software program for the Robotic Fight Car (RCV) program.


SITE AD for the 2025 Robotics Summit registration.
Register now so you do not miss out!


Scalable ML & Information Testing


In as we speak’s digital platforms-from purchasing apps and streaming companies to well being trackers and buyer portals-machine studying is central to how methods personalize experiences, automate selections, and reply to person actions. However irrespective of how superior a mannequin is, it may possibly fail if the info feeding it isn’t dependable.

Creator: Naga Harini Kodey, https://www.linkedin.com/in/naga-harini-k-3a84291a/

With fixed streams of person interactions-clicks, swipes, logins, transactions, and events-flowing by way of these methods, sustaining information accuracy turns into a foundational requirement. Damaged information pipelines, inconsistent function values, and unmonitored modifications can result in silent failures. These failures typically go unnoticed till person satisfaction drops or key enterprise metrics take a success.

As a Principal QA Engineer, I’ve collaborated carefully with engineers, analysts, and information scientists to check machine studying pipelines end-to-end. This text outlines sensible QA methods and hands-on methods that may be utilized throughout platforms pushed by real-time or batch person information, serving to groups stop points earlier than they affect manufacturing.

The place Issues Go Fallacious in ML Pipelines for Consumer Programs

Consumer-driven platforms acquire information from a variety of sources-web exercise, cell apps, sensor inputs, and exterior APIs. As this information flows by way of ingestion, transformation, and mannequin scoring, there are a number of widespread failure factors:

  • Lacking fields in logs → Instance: System sort or session ID not logged constantly throughout cell and internet.
  • Inconsistent occasion naming → Instance: checkoutInitiated modified to checkout_initiated, breaking downstream dependencies.
  • Unrealistic or incorrect values → Instance: Session time reveals zero seconds or logs a person clicking 200 occasions in a second.
  • Code modifications with out validation → Instance: Characteristic transformation logic up to date with out verifying downstream mannequin compatibility.
  • Mismatch in coaching vs. manufacturing → Instance: Fashions educated on curated information however deployed on noisy, real-world inputs.
  • Check site visitors contaminating dwell information → Instance: Automated testing scripts inadvertently included in manufacturing metrics.
  • Damaged suggestions loops → Instance: Retraining logic relies on a sign that silently stops firing.

These issues typically degrade efficiency subtly-skewing suggestions or altering person flows-making them more durable to detect with out focused validation.

Scalable ML & Information Testing

Testing Methods That Work in Observe

Every stage of the pipeline-from uncooked occasion seize to function transformation to mannequin output-presents a novel testing alternative. Right here’s a breakdown of sensible methods:

1. Begin on the Supply: Uncooked Information Validation

Widespread points: Lacking timestamps, corrupted machine IDs, inconsistent information codecs.

How one can check it:

  • Construct schema validators utilizing instruments like Nice Expectations or Cerberus.
  • Set automated thresholds for lacking values (e.g., alert if >5% of user_id fields are null).
  • Monitor ingestion volumes over time; flag sudden drops/spikes in key occasions.

Instance Implementation:

python –

assert occasion[‘timestamp’] is just not None

assert isinstance(occasion[‘device_id’], str)

2. Confirm Characteristic Logic

Widespread points: Incorrect logic in options like session length, or loyalty rating.

How one can check it:

  • Write unit checks for transformation capabilities utilizing identified pattern inputs.
  • Outline worth bounds or anticipated distributions (e.g., session length shouldn’t be > 12 hours).
  • Embrace logging checkpoints to confirm computed values at every stage.

Guidelines Tip: Create a function contract doc itemizing every function, supply columns, transformation steps, and check circumstances.

3. Look ahead to Coaching vs. Manufacturing Drift

Widespread points: Characteristic values differ between coaching and manufacturing environments.

How one can check it:

  • Run statistical comparability (e.g., KS check or PSI) between offline coaching information and dwell enter information.
  • Add a nightly job to check means, medians, and ranges of energetic options.
  • Visualize function drift on dashboards to trace gradual degradation.

Alert Instance: “Characteristic X imply has shifted from 0.2 to 0.45 over the previous 7 days.”

4. Lock Down Enter and Output Expectations

Widespread points: Schema mismatches, renamed fields, or lacking inputs trigger the mannequin to misbehave.

How one can check it:

  • Use golden input-output pairs as regression circumstances in your CI pipelines.
  • Add an enter validation layer that enforces construction, information varieties, and presence of required fields.
  • Log and evaluate mannequin output distributions throughout variations.

Observe Tip: All the time pin a “canary” check with a identified document that ought to give a set prediction rating.

5. Monitor for Silent Failures

Widespread points: All the pieces runs, however person engagement or conversions drop unexpectedly.

How one can check it:

  • Construct dashboards for monitoring scoring quantity, function completeness, and mannequin predictions.
  • Cross-check enter function presence day by day and evaluate it with coaching schema.
  • Arrange anomaly detection on output KPIs (conversion price, engagement price).

Instance: “If purchase_probability output from the mannequin drops by 30% over 3 days, flag it for investigation.”

Finest Practices for Testing ML Pipelines

  • Check early, check small: Validate information earlier than it hits your transformation logic.
  • Create edge circumstances: Deliberately go invalid or boundary values to check mannequin resilience.
  • Monitor and model every part: Keep lineage for datasets, options, and scripts.
  • Automate regression checks: Each mannequin launch ought to be backed by automated state of affairs validation.
  • Collaborate throughout capabilities: QA, information science, product, and engineering ought to evaluation pipelines collectively.
  • Make failures seen: Spend money on real-time alerting and dashboards. Fewer surprises = higher outcomes.

Conclusion

For platforms pushed by person interplay, machine studying can’t succeed with out reliable information. When pipelines break silently, the affect hits person expertise, retention, and income. Testing these methods must be proactive, systematic, and tailor-made to real-world circumstances.

Scalable check protection ensures each component-from information ingestion to mannequin scoring-holds up beneath stress. By specializing in root-level information integrity and transformation validation, QA groups grow to be vital gatekeepers of efficiency and reliability.

Testing isn’t nearly catching bugs-it’s about safeguarding the intelligence behind your platform.

References / Additional Studying

Concerning the Creator

Naga Harini Kodey is a Principal QA Engineer with over 15 years of expertise in automation, information high quality, and machine studying validation. She focuses on testing AdTech information pipelines and ML workflows, builds check frameworks, and a worldwide speaker on QA methods, information testing and end-to-end machine studying system assurance.