Meet your new colleague – the ML Admin, who tames LLMs so they're ready to rock
- Reference: 1747286948
- News link: https://www.theregister.co.uk/2025/05/15/new_job_machine_learning_administrator/
- Source link:
“It’s a new persona, but not a well-known persona,” Debbo Dutta, Nutanix’s Chief AI Officer told The Register at last week’s Next conference.
“The ML admin articulates, designs, plans, executes, and monitors the emergent large language model lifecycle,” Dutta said.
[1]
The gig doesn’t apply to the tech used to train models, Dutta said. Instead this new class of admin will need to understand how to size large language models to match them with infrastructure and workloads, and ensure uptime for models.
[2]
[3]
They’ll also need to be across compliance matters, understand infrastructure well enough to scope it and work with ops teams to build it.
“Our hypothesis is that the IT admin will continue to manage the infrastructure up to the layer of the large language models. That’s the red line for now,” Dutta said.
[4]
ML admins will also manage their employer’s corporate accounts with major AI vendors such as OpenAI and Anthropic. Doing so is necessary because as application vendors add AI to their products, they’ll assume users have the required large language model in place. The ML admin’s job will be to provision their employer’s AI resources so applications can put them to work.
[5]NSF director memo to staff: Don't worry about those job cuts, at least for now
[6]Bosses weren’t being paranoid: Remote workers more likely to start own biz
[7]IT pros are caught between an AI rock and an economic hard place
[8]AI agent promotes itself to sysadmin, trashes boot sequence
“Our hope is that very soon ML Admins will be part of IT,” Dutta said. For now they mostly work in business units - yet another example of shadow IT emerging as business units build their own AI apps without IT department approval and hire staff to keep them humming.
ML Admins may need to be grown, not hired.
Dutta said Nutanix currently has a “shadow ML Admin, who is training an It admin how to be an ML Admin.”
That means Nutanix is building out its capability to run and manage AI even as it builds that AI. Or as Dutta put it: “We are building the runway, and the control system, and flying the plane at the same time. It is a little crazy right now.”
[9]
But at least it’s a new job related to AI at a time the tech is [10]contributing to layoffs. ®
Get our [11]Tech Resources
[1] https://pubads.g.doubleclick.net/gampad/jump?co=1&iu=/6978/reg_software/aiml&sz=300x50%7C300x100%7C300x250%7C300x251%7C300x252%7C300x600%7C300x601&tile=2&c=2aCW7QV6-MsYpXT5Ifr11gQAAAYA&t=ct%3Dns%26unitnum%3D2%26raptor%3Dcondor%26pos%3Dtop%26test%3D0
[2] https://pubads.g.doubleclick.net/gampad/jump?co=1&iu=/6978/reg_software/aiml&sz=300x50%7C300x100%7C300x250%7C300x251%7C300x252%7C300x600%7C300x601&tile=4&c=44aCW7QV6-MsYpXT5Ifr11gQAAAYA&t=ct%3Dns%26unitnum%3D4%26raptor%3Dfalcon%26pos%3Dmid%26test%3D0
[3] https://pubads.g.doubleclick.net/gampad/jump?co=1&iu=/6978/reg_software/aiml&sz=300x50%7C300x100%7C300x250%7C300x251%7C300x252%7C300x600%7C300x601&tile=3&c=33aCW7QV6-MsYpXT5Ifr11gQAAAYA&t=ct%3Dns%26unitnum%3D3%26raptor%3Deagle%26pos%3Dmid%26test%3D0
[4] https://pubads.g.doubleclick.net/gampad/jump?co=1&iu=/6978/reg_software/aiml&sz=300x50%7C300x100%7C300x250%7C300x251%7C300x252%7C300x600%7C300x601&tile=4&c=44aCW7QV6-MsYpXT5Ifr11gQAAAYA&t=ct%3Dns%26unitnum%3D4%26raptor%3Dfalcon%26pos%3Dmid%26test%3D0
[5] https://www.theregister.com/2025/05/14/national_science_foundation_memo/
[6] https://www.theregister.com/2025/05/12/remote_work_leads_to_more_startups/
[7] https://www.theregister.com/2025/05/05/between_the_ai_rock_and/
[8] https://www.theregister.com/2024/10/02/ai_agent_trashes_pc/
[9] https://pubads.g.doubleclick.net/gampad/jump?co=1&iu=/6978/reg_software/aiml&sz=300x50%7C300x100%7C300x250%7C300x251%7C300x252%7C300x600%7C300x601&tile=3&c=33aCW7QV6-MsYpXT5Ifr11gQAAAYA&t=ct%3Dns%26unitnum%3D3%26raptor%3Deagle%26pos%3Dmid%26test%3D0
[10] https://www.theregister.com/2025/05/13/microsoft_layoff/
[11] https://whitepapers.theregister.com/
"ML Admins may need to be grown, not hired."
Well... there's probably enough manure around this whole "A.I." thing to do just that...
ML Admins may need to be grown, not hired.
Almost a canned BoFH episode in that sentence alone.
Grown from what? The spent seed of the C-Suite tossers?
Most IT admin roles reduce to resource assessment, allocation and management including performance tuning. So what particularly is so very different about the administration of LLMs. Running a model with a stream of queries is not too dissimilar to running at database system at that level I would have thought.
'Special Sysadmin then
so it's someone who has drunk the koolaid and actually likes LLM unlike most SA who know to stay well away so the stink doesn't get on them... That's the only reason it's a separate role.
Anyone else
Bored of the continually re-hyping of AI/ML LLMs etc?
I had nearly got through this week without seeing some shite about ML being awesome too!
"Manage compliance"
Keeping your LLMs compliant with pretty much anything when they make up answers, it is trivially easy to trick them into bypassing security, they leak whatever data is passed through them to their progenitor companies, and nobody understands what they are actually doing, does sound like a full-time job.
Applications with added AI?
>Doing so is necessary because as application vendors add AI to their products, they’ll assume users have the required large language model in place.<
I thought the idea is you do not have to change apps for AI, just just give the LLM access to all the data you have and the AI will then do everything you tell it to do with a simple prompt.
How would an application with added AI make use of a LLM a company already has in place? Where is the system architecture and system interface to that?
Did Nutanix provide some documents on how they think this is gonna work or is this a dream of Chief AI officer Dutta or the output of their AI?
Bingo!
> The ML admin articulates, designs, plans, executes, and monitors the emergent large language model lifecycle
Any more buzzwords we could stuff into this sentence?
I can see the need for an MLA, just as I understand the need for a DBA, network admin and all the other specialised roles. The question that comes to mind is why this would have to be a person. Isn't the entity best placed to monitor LLMs and all the rest of it just another LLM.
In fact I can foresee an entire shadow organisation appearing within IT companies. One that is made up entirely of AIs, agents, MLAs and a few yet-to-be-created roles in AI management. I just wonder how much of their time will be spent in [1]pointless machine to machine meetings
That should slow 'em down to the point of uselessness.
[1] https://www.techradar.com/computing/artificial-intelligence/two-ai-chatbots-speaking-to-each-other-in-their-own-special-language-is-the-last-thing-we-need