Legislate meets TurinTech, the company empowering businesses to build efficient and scalable AI

Why sustainable AI starts with optimisation

Create contracts on no legal budget

Get started

In this episode, Legislate meets Dr Leslie Kanthan, CEO and co-founder of TurinTech. Leslie shares how TurinTech is making machine learning models more sustainable thanks to model and code optimisation and provides some AI use cases ranging from finance to legal services.

Listen to the episode below:

Learn more about TurinTech

Learn more about Legislate

Listen to more episodes

All Episodes

Create a contract for £9.95 or Subscribe Monthly

View Plans

Read the transcript

Charles Brecque: Welcome to the Legislate podcast, a place to learn about the latest insights and trends in property, technology, business building, and contract drafting. Today, I'm excited to welcome Dr Leslie on the show, CEO and co-founder of TurinTech. TurinTech enables businesses and organisations to build efficient and scalable AI at speed through automating the whole data science life cycle. Leslie, welcome to the show. Would you like to introduce yourself and share a bit of background about TurinTech?

 

Dr Leslie Kanthan: I'm the CEO of TurinTech. I have a background in mathematics, specialising in graph theory. I did a masters, as well, in that area, and then continued to have a PhD and post-research background, as well, in machine learning using graph-based methods. I've worked in the financial services industry, for banks, for many years. Here I am now, with Turintech, which is a company that specialises in optimisation for machine learning models. I founded the company with 3 other co-founders, also with very similar backgrounds, friends who worked in the financial services industry, and we're solving the kinds of problems that we used to come into contact with, day in and day out, using AI and automation.

 

Charles Brecque: When you say 'optimising AI', how do you go about doing that?

 

Dr Leslie Kanthan: We have a very famous research paper on code optimisation using very specific AI techniques. It got into a top tier conference, and it was one of the very few papers that we released, not kept proprietary secrets. What we do is we optimise the underlying code. Essentially, machine learning models can be broken down into code, and we perform the optimisation at the code level. We also do optimisation using parameter optimisation, and a lot of other techniques that encompass that end-to-end. That's what we do.

 

Charles Brecque: If I were a customer, why would I want to optimise my machine learning models?

 

Dr Leslie Kanthan: Now, with the advent of big data, and volume of data, data scientists emerging, a lot of data scientists need tools so that they can augment their capability. A data scientist's job is to generate models, and the business analyst's job is to use those models, to get insights from those models. A trader's job is to use those models to develop profitable trading strategies. A lawyer's job is to use those models to see if he can match some of the text data from the file to the correct fee earner. Everybody's using models. It's not just in the financial services or tech industry. We essentially make those models more accurate, and more efficient. One of those key areas is in terms of sustainability. To generate models is a very computational-intensive process, and that can be very costly, in terms of GPUs and electricity consumption, etc. Our optimisation, and code optimisation which we have showcased in one of our earlier papers-, we scientifically proved that we can optimise code, reducing the energy consumption of the device, or anything that's running that code. That's where the energy cost savings can be made, in particular to the process of what we're doing.

 

Charles Brecque: As an entrepreneur, what's been your favourite moment so far?

 

Dr Leslie Kanthan: I think my favourite moment is having the scientific know-how around us. We have some of the best machine learning scientists, PhD level, some of them professor level, working with us to build out our IP and effectively make our optimisation more and more accurate, and more and more efficient. That's one of those moments for me that I feel that having all of these guys working around us, and sharing these ideas, and sharing the knowledge, and having a lot of fun with them, that's great.

 

Charles Brecque: What do you wish you had known before starting TurinTech?

 

Dr Leslie Kanthan: I wish that I knew that there's more to artificial intelligence than what we initially thought. There are a lot of people who may not have a technical background, or an AI background, who are very interested in learning more about AI and the applications. If I knew that beforehand, then we could have had earlier discussions, and got to networking, and know more about those use cases earlier on.

 

Charles Brecque: Where do you see your business in 3 to 5 years?

 

Dr Leslie Kanthan: You have Microsoft Office for documentation processing. You have DocuSign for documents. You have Tableau for charting. What do you really have for AI? I think our vision for our business is to be exactly that for all your AI needs, and if you need to develop AI software then you use our product. That's where I see the vision of us being more of an AI-tooling software solution, a complete environment, in the future.

 

Charles Brecque: As a CEO, I imagine you interact with key contracts quite a bit. What can you share about them?

 

Dr Leslie Kanthan: Everything from NDAs to POCs, statement of works for software vendor, preferred supply lists. There's a lot of contracts involved in a lot of processes. I have quite a lot of interaction with them.

 

Charles Brecque: With those contracts, are there any areas of friction that you've encountered, that you've had to overcome, and how did you overcome them?

 

Dr Leslie Kanthan: I think contract processes are very long, in general. They can be many pages, and that requires a lot of interaction from many different people from the team. It can involve somebody from the tech side, most often the legal side, so interacting with our lawyers. We have a good set of lawyers to help us with that. There's a lot of back and forth to understand some of the terms, and see if there are any terms that we want to change and propose therefore.

 

Charles Brecque: It seems like you use lawyers to overcome those potential bottlenecks?

 

Dr Leslie Kanthan: Who doesn't use lawyers? Of course, yes.

 

Charles Brecque: I would say Legislate users don't use lawyers, but the type of contracts that we offer are relatively standard agreements, which you could use a lawyer for, but what we offer are lawyer-proof contracts that can be tailored, within reason.

 

Dr Leslie Kanthan: That makes sense. I think that Legislate should be using our product then. If you have AI, and you can combine that, then you can essentially read some of the text, and do a lot of the customisation using natural language processing to adapt to the needs of a potential client, and then avoid, or limit, the need for a lawyer.

 

Charles Brecque: We've taken a knowledge graph approach to configuring the contracts, so we're encoding the logic that a lawyer might apply to create and tailor an agreement, and we're applying those rules to the parameters of the contract, to build a contract. There isn't necessarily much optimisation of that aspect. We know what the templates are, in the sense that we've sourced them from gold standard libraries. We know what the different options are, and we're not necessarily trying to change the wording of the clauses, because it's all relatively standard, and we're trying to keep it standard for our user group. I think, in terms of where there's a lot of potential, we have over 2,000 members creating contracts on Legislate, and there's a lot of user data that we're gathering. What we're trying to do is aggregate that data into useful statistics for people who are not used to creating contracts, so that they know what the average, the norm looks like. Where we could use machine learning, also, is to forecast signature outcomes. 'How long will it take my contract to be signed if I choose this set of terms versus another set of terms?' There'll be lots of optimisation to do.

 

Dr Leslie Kanthan: Yes, interesting.

 

Charles Brecque: I'm conscious, Leslie, that we've taken a lot of your time already, so I'm going to ask you the closing question we ask all our guests. If you were being sent a contract to sign today, what would impress you?

 

Dr Leslie Kanthan: If it's simple, and if it's substantially rewarding.

 

Charles Brecque: Then you have the best of both worlds.

 

Dr Leslie Kanthan: It seems so.

 

Charles Brecque: Thank you very much, Leslie, for your time, and best of luck optimising those models, and hopefully Legislate one day.


Most recent Articles

Go to Blog

Most popular contracts

Go to Contracts