Kirkland Hints It Could Fine-Tune LLMs For Own Legal AI Model

Kirkland & Ellis’s $500m tech project could involve fine-tuning open source LLMs to create their ‘own’ legal AI model, their hiring binge for innovation roles has suggested. Artificial Lawyer found that two new AI Infrastructure Director positions demand experience with ‘on‑premise GPU environments’, i.e. the skills you would need if you wanted to fine-tune an open source LLM to meet your specific needs – as Thomson Reuters is currently doing (see here.)

The scope of these new roles, and many others listed on their recruitment site, help to show what is planned as part of the massive DIY project to move away from what it calls the legal AI ‘floor’, i.e. the off the peg platforms.

Although, in some of its job adverts for the project they mention they want people to have experience with ‘Harvey, Legora, CoCounsel, Lexis+ AI’. And that makes sense. If you plan to build your own internal legal AI platform then what better examples to base your designs on?

Back to GPUs.

There are two new roles with the title AI Infrastructure Director, posted May 27 – just around the time of the FT announcement, one in Huston and one in Chicago, with salaries from $302,000 to $335,000, if you’re interested. Here are some key aspects of the role.

• ‘As AI Infrastructure Director, you’ll design, manage, and optimize the firm’s AI infrastructure – spanning on‑premise GPU environments and Microsoft Azure–based AI platforms – to enable enterprise AI, automation, and innovation initiatives at scale.’

And note, it doesn’t say ‘you have some experience with GPUs in general’, it says ‘optimize the firm’s AI infrastructure – spanning on‑premise GPU environments’. I.e. that’s quite specific in terms of who owns the GPUs in this scenario.

Now, you might want GPUs for something else entirely, but in this context it’s not a reach to suggest they’d be used to help with fine-tuning an open source LLM with your own data. And a firm such as Kirkland would have a LOT of data. And, see below, they’re also hiring – and have been hiring already – for a ton of roles around AI use and workflow mapping.

Meanwhile, in terms of the Infra role, they added it covers:

• ‘AI Infrastructure Ownership: Lead all AI environments—including on‑premise Graphics Processing Unit (GPU) clusters, Microsoft Azure AI and Machine Learning (ML) services, and shared AI platform components—with accountability for reliability, scalability, and lifecycle management.
• Innovation Enablement: Create secure, governed environments that enable rapid experimentation and development for Innovation, AI Engineering, and CGO teams.
• Platform Design & Delivery: Oversee the design and deployment of shared and custom AI platforms that accelerate solution delivery while meeting security and governance standards.’

And here are some more indicators. i.e. they mention the need for ML services, i.e. training, and they talk about ‘governed environments for experimentation’ – that’s a phrase right out of the builders’ manual.

We’ll get into the ‘what does it mean?’ aspects in a moment. The Kirkland jobs site lists around 85 roles that mention AI, some of which were listed as far back as March, with several added every few days.

They’ve said that they will use a team of around 180 people to achieve the massive DIY project, but clearly many roles are still being recruited for, and older job adverts are still posted on their own recruitment pages, suggesting they are still looking for people, or additional hires, for those listed roles.

This other job also spells things out very clearly: AI Innovation Adviser, with a regionally-adjusted pay range from $153,000 to $220,000. And they are hiring – or have already hired – for several people in this role across the US.

Here are some of the key aspects:

• ‘As an AI Advisor you will embed within practice groups to translate legal tasks and workflows into scoped AI solutions,
• [you will] own end-to-end workflow development, partnering with engineers, and contributing to a scalable playbook of best practices.
• In addition to internal firm initiatives, this role may support client-facing matters, partnering with practice teams to deliver AI-enhanced legal services directly to clients. 
• You will also champion responsible AI adoption, designing training that empowers attorneys to use these tools with confidence.
• The AI Advisor sits within the Innovation Program AI Advisory Services team.
• The ideal candidate is equal parts trusted advisor and hands-on builder, someone who thrives on rapid experimentation.’

And, as to what you need to get this role, you need:

• ‘A J.D. from an accredited law school is strongly preferred; however, candidates who do not hold a law degree but possess five or more years of demonstrated experience leading innovation initiatives.
• Strong aptitude for process improvement and legal technology, including generative AI/large language models with sufficient understanding of where AI succeeds and fails to engage both legal teams and engineers.
• Direct experience with legal AI platforms (Harvey, Legora, CoCounsel, Lexis+ AI, or comparable), including hands-on building and iterating on prompts and workflows with accountability for output quality.
• Strong writing, project management, and organizational skills with sound judgment.’

—

(conference advert)

Legal Innovators California is in just over one week, and we have an incredible group of speakers taking part from across the legal AI ecosystem – plus we’ll be right in the heart of the Bay Area, the home of the AI world! Join us for the West Coast legal event of the year on June 10 and 11 in San Francisco!

Tickets For Legal Innovators California Conference – SF – June 10 + 11

To gain your complimentary pass for the two-day landmark legal AI event in San Francisco, please use the Express Registration here – only for those working within law firms or inhouse teams.

—

What Does This Mean?

Well, AL did wonder how they’d spend that $500m, given they already will have a tech stack that goes from the basic Word suite to a range of major legal research tools. But, fine-tuning an LLM with your own GPUs does show where some of the cash could potentially go.

The many other open roles for legal tech folks to work hand in hand with practice groups to model their workflows will also absorb a lot of time and cash. (And, as an aside, it reminds AL a bit of what Eudia is doing via its digital twins strategy for inhouse lawyers.)

It also shows that they are serious. At first it seemed that this might just be a very expensive marketing campaign, but without a British film star involved. But, if you look at all the job specs then they truly are serious in intent. The details show really well-developed plans and the scale of hiring underlines they mean what they say.

Will it overturn the status quo of buying off the peg? Probably only for a few firms. As seen, this is a major project and not every firm has the spare cash to do this. Also, although Kirkland will no doubt build something that works for them, will it be head and shoulders above what they could have done with third party tools, plus some customisation, plus vibecoding some extra tools as and when needed? It’s hard to say, as we have not seen the finished product yet.

Will clients really see a radical difference? Again it’s hard to say. Would a Latham & Watkins or a Skadden, with a very deeply embedded set of third party platforms and some great DIY point solutions, not provide generally the same outputs….at least as far as clients could tell? Again, we will have to wait and see what comes from this.

One additional point to add is around data privacy. Even if the actual outputs are only marginally better than using customised third party tools, what Kirkland could potentially have with their own system, especially if it’s running off their ‘own’ fine-tuned open source LLM with an ‘on prem’ set-up, is increased privacy.

The additional privacy aspect could perhaps be seen as an important differentiator. Of course, with enterprise solutions of all kinds there is also data security built in. But, having your own legal AI system on premises does at least send the message that you have gone the extra mile.

Interesting times and kudos to Kirkland, because whatever happens next they have underlined to the legal sector the importance of legal AI – and for Artificial Lawyer that’s definitely a positive!

—

(conference advert)

Legal Innovators Europe Conference in Paris

The in-person, two-day legal tech event in Paris takes place on June 24 and 25. If you are at a law firm or inhouse you can gain a free pass via the Express Registration below.

Legal Innovators Europe conference in Paris – June 24 (Law Firm Day) and 25 (Inhouse Day).

Express route to your ticket here.