EnterpriseAI

Deploying AI in the Enterprise
Join this virtual event and interactive discussion with industry experts about AI and data-driven digital transformation in the enterprise. Learn how HPE (with the recent BlueData acquisition) is helping accelerate time-to-value with AI, machine learning, and advanced analytics.
Top Votes
Peter Burris
How do you address the lack of skilled data scientists, engineers, and analysts? What can be done to help improve productivity and collaboration across data science teams? https://www.crowdchat.net/s/25tgs
https://www.crowdchat.net/s/25tgs

[+] Show Hidden Comments
[-]
ingrid (vdh) burton
We have H2O Driverless AI that automates machine learning for the enterprise. It helps scale an existing data scientist or data engineer to arrive at results.
[-]
contriveit
Partnering can be a quick answer to the lack of skills. That my daily job, work with customers who don't have all the tools and skills to do it for themselves.
[-]
jameskobielus
Use visual tooling that lets subject matter experts build declarative logic that drives the autoML CI/CD workflow that models, trains, and serves AI/ML assets/apps that meet business requirements and drive desired outcomes.
[-]
Tom Phelan
Tools such as H20 are helping with this
[-]
Frederic Van Haren
Using an efficient ecosystem of tools to reduce and compensate the need for skilled resources.
[-]
Manoj Suvarna
Investing in the right #AI tools and partnering with #HPE who can be your #trusted #advisor with #services to assist our customers in accelerating their #AI Business outcomes
[-]
Mike Leone
Organizations will try to hire internally or leverage management consultancies, but professional services will be essential to guide organizations through the AI journey. 38% of organizations will turn to vendor-provided professional services and consulting.
[-]
ingrid (vdh) burton
We also need to get more people trained globally. We are working with small data science teams from Kuwait, South Africa, Paraguay, Chili, and various other countries. There is a huge need out there - across the globe.
[-]
Manoj Suvarna
#People #Technology #Partner #Ecosystem
[-]
contriveit
In the longer term clearly more skills are needed to maintain the pace of adoption. We are working with some great graduate programs to give industry experience to those fresh people.
[-]
NandaVijaydev
Data Science requires a broad spectrum of skills. It is not an individual sport. So lack of skills can be addressed by making the experts share their ideas, models, code, and docs, with the less experienced ones.
[-]
Peter Burris
Will we see "data science-as-a-service" anytime soon?
[-]
Frederic Van Haren
@jameskobielus Yes, no need to re-invent the wheel.
Peter Burris
In your experience, what are some of the challenges that data science teams face in deploying and operationalizing distributed AI / ML and advanced analytics in a large organization? https://www.crowdchat.net/s/45teb
https://www.crowdchat.net/s/45teb

[+] Show Hidden Comments
[-]
Frederic Van Haren
Researchers don't care about the underlying HW as long as they can run their workloads on their favorite platform in a timely and efficient way.
[-]
Tom Phelan
Common challenges are cost control, reliable reproduction of cluster configuration, data and cluster security.
[-]
Patrick Osborne
#dirtydata
[-]
ingrid (vdh) burton
We see three challenges: Talent, Time, Trust.
[-]
Hande Sahin-Bahceci
I like your 3 Ts
[-]
ingrid (vdh) burton
Talent - there aren't enough data scientists today. So providing a platform that augments what they are doing is necessary.
[-]
Frederic Van Haren
Lack of complete solutions that satisfy the business needs.
[-]
Storage Godfather (HPEStorageGuy)
ESG had a data point just in the last week I saw that said 38% of orgs investing in AI initiatives are doing so without a data scientists. That was shocking to me.
[-]
Mike Leone
skills shortages in other areas of the data pipeline. Data scientists are spending more time on completing tasks associated with data integration and data preparation rather than the data science tasks like modeling, training, tuning, etc.
[-]
ingrid (vdh) burton
Time: It can take months to do modeling. We can cut that time to weeks or hours with automatic ML on BlueData
[-]
NandaVijaydev
Data quality, duplication, availability, lack of compute resources to process larger datasets, lack of consistency while building and deploying models
[-]
contriveit
Getting into production. Its great developing into a closed environment but data scientists and analysts don’t want to need to worry about the operational side. Anything to ease that rollout is a huge bonus right now.
[-]
Manoj Suvarna
In most cases, Data Scientist resources end up spending nearly 80% of their time sourcing and cleaning the data while also optimizing on the right #infrastructure to deploy their models
[-]
Don Wilson
- One of the big challenges we hear from customers is the project objectives often change mid-stream. Implication is that the ecosystem must be designed so that it can accommodate those dynamics
[-]
Peter Burris
@patrick_osborne And ignorance that the data is dirty!
[-]
Patrick Osborne
I agree with @fvha in that it is our responsibility as IT professionals to build a scalable service so data scientists and data engineers only worry about their tools and data sets.
[-]
ingrid (vdh) burton
Trust: Data scientists have to explain their results. They need to be able to say to a regulator: here's how I got the answer. We do that with Machine Learning Interpretability... transparency to reveal your results.
[-]
Frederic Van Haren
A lot of time can be spent putting together and maintaining an AI environment. It is costly, time consuming and to a certain degree it is reinventing the wheel. Time is better spent on solving the problem at hand.
[-]
Hande Sahin-Bahceci
Agreed, is it lack of resources or lack of aligning these resources to the innovative areas vs. operational tasks?
[-]
Tom Phelan
agreed. the use cases are dynamic
[-]
jameskobielus
One of the chief operationalization challenges for AI/ML apps is ensuring that the models, wherever developed (TensorFlow, PyTorch, etc.) are compiled to the most efficient format for downstream inferencing.
[-]
contriveit
@itcontrive Don’t forget, the models are only a small part of the value-chain. You need to feed the models and act upon the answers they give you. If you don’t act in a timely fashion, you might well have never made the prediction.
[-]
Peter Burris
@CalvinZito What do you think that 38% doing AI without data scientist means? Are they depending on external service providers?
[-]
Ralph Finos
understanding the business issues and business language so the results are on-point and actionable. large organizations are iikely to have competing internal interests and views that must be translatable and accommodate relevant stakeholders..
[-]
ingrid (vdh) burton
@fvha Correct! They are in the business of gleaning answers and insights.... not worrying about the underlying infra is a time saver!
[-]
Patrick Osborne
@msuvarna do you see an opportunity for HPE to at least mask the complexity of managing their infrastructure?
[-]
ingrid (vdh) burton
@Hande27 Thanks!
[-]
Storage Godfather (HPEStorageGuy)
I would hope so - don't see how you can successfully do an AI project without data scientists engaged in it.
[-]
jameskobielus
Another key challenge for AI/ML DevOps in large orgs is ensuring unified CI/CD that's aligned with code governance across disparate app-dev initiatives. Wikibon is seeing more enterprises incorporate "big data catalogs" to underpin that requirement.
[-]
Hande Sahin-Bahceci
IT&Ops side of AI/ML deployments can be abstracted and automated for data science teams. That is why our #HPE Pointnext data scientists get excited about BlueData.
[-]
Lawrence Hecht
@jameskobielus Unified CI/CD is too much to ask. Traditionally, there has not been an attempt to standardize CI tools across the entire enterprise
[-]
Lawrence Hecht
@CalvinZito you don't need data scientists to use "AI" if you are just assuming it is built into other software you're using.
Peter Burris
What kinds of business value and business impact are you seeing for enterprise AI / ML and advanced analytics? For what kinds of use cases? https://www.crowdchat.net/s/05tdh
https://www.crowdchat.net/s/05tdh

[+] Show Hidden Comments
[-]
ingrid (vdh) burton
In almost every industry whether Finserv, healthcare, manufacturing, retail... customers are getting real results with AI and ML.
[-]
Frederic Van Haren
The democratization of AI means that AI is not just a platform for research but also a platform for mainstream Corp IT. This tells CIOs to look at AI as an additional service they need to provide.
[-]
Manoj Suvarna
Customers may be seeing increased profitability, reduction in costs or improved customer experience using AI/ML!
[-]
Don Wilson
Sandbox as a Service, GPU IaaS
[-]
Patrick Osborne
As an example within HPE, we are using AI/ML within our Infosight platform to provides AIOps to assist and provide predictive analytics for our platforms. Case reduction of incidents by over 90+%
[-]
contriveit
Lots of interest based around predictive and prescriptive maintenance. Equally as much interest on image analytics and the pipelines they need. Manufacturing seems to be leading the charge.
[-]
Storage Godfather (HPEStorageGuy)
@patrick_osborne And ultimately, happier customers.
[-]
ingrid (vdh) burton
Use cases include: better credit scoring, credit risk, fraud detection, KYC in Healthcare: Sepsis detection, saving lives, matching patients to doctors for faster treatments.... the list goes on and on
[-]
jameskobielus
Enterprise AI/ML and advanced analytics have a broad scope of use cases. Let's start with the obvious: customer and marketing analytics to drive loyalty, upsell, satisfaction, and experience. All the predictive analytics against customer data.
[-]
Mark Hitchcox
In FSI we're seeing AI & ML being used for the purposes of Fraud detection & prevention. Leveraging well known analytics with ML in real time to alert & prevent fraud at the institutional, transnational & customer level .
[-]
Tom Phelan
Predictive analytics also being used in healthcare and financial industries
[-]
ingrid (vdh) burton
Absolutely! Banks are seeing a greater propensity to sell credit products. Doctors and patients are seeing better outcomes. Delighting customers is absolutely key.
[-]
Manoj Suvarna
Use cases for #AI/ML vary from predictive maintenance to fraud detection to quality control to customer sentiment analysis...
[-]
Patrick Osborne
we have customers not only using AI/ML, but with thie ability to quickliy stand up models, they are able to improve results rapidly, retrain and redeploy using BlueData
[-]
NandaVijaydev
Healthcare has a number of use cases. Patient care, improving hospital supply chain, long term care and more
[-]
Frederic Van Haren
Non-traditional verticals such as HR are seeing benefits from AI.
[-]
jameskobielus
Now let's talk about the more advanced AI use cases: face, speech, natural language, and other data-driven recognition and understanding use cases. In other words, automating human-like cognition for everything, such multifactor authentication from your smartphone.
[-]
contriveit
In government we are seeing lots of organisations seeking to do more with less. Make those budgets stretch further.
[-]
ingrid (vdh) burton
Many of our customers are using Automatic Machine Learning to get business outcomes... We are seeing new answers to problems that businesses have.
[-]
Mike Leone
The most important objective at this point is increasing operational efficiency and the low hanging fruit is within IT, system management, orchestration functionality, log file analysis, etc.
[-]
Jason Schroedl
the business impact for AI / ML often falls into a few different categories: maximizing operational efficiency; improving the customer experience; reducing risk; and delivering innovation with a new business model or new services / product
[-]
Patrick Osborne
I would love to use a wearable devices and never have to go to the doctor again, they would proactively call me!
[-]
David Floyer
AI applications are still difficult to develop, test and deploy. Success seems to be connected to good unique access to unfiltered data.
[-]
Hande Sahin-Bahceci
Better asset management with prescriptive maintenance, quality assurance with video analytics, efficient parking with smart city are just few AI use cases HPE Pointnext experts delivered.
[-]
jameskobielus
Interestingly, AI/ML has been well entrenched in anti-fraud, credit scoring, IT operations management for many years. Using ML to sift through petabytes searching for the anomalous patterns that need prompt remediation/response.
[-]
David Floyer
@patrick_osborne Agreed! Real-time analytics at the edge combined with contextual history could revolutionize healthcare.
[-]
Lawrence Hecht
@patrick_osborne That seems like a crazy high number. Are we talking about "serious" incidents?