Today’s guest is Audrey Smith, Chief Operating Officer at MLtwist in Santa Clara, CA. Founded in 2021, MLtwist enables you to seamlessly connect with their partner’s labeling and workforce while making sure that your data is safe and securely formatted and classified. Their integrations will push your data assets to the best platform for your specific use cases and send it back in the format that your Machine Learning team requires.
Scientists and engineers leverage MLtwist’s deep Labeling Operations expertise, live Data Labeling Platform marketplace, real-time dashboards and robust integrations with 75+ integrated data labeling platforms. Their services take care of the time consuming complexities that go into data labeling operations – including designing workflow, testing the right data labeling platforms, writing guidelines & training workforces – all the way to quality control.
In the episode, Audrey will discuss:
– MLtwist’s work within data preparation & labeling
– Benefits their DaaS and SaaS offerings bring to customers
– The team structure and unique role in Project Management
– Why to consider a career with MLtwist
– What the near future looks like for MLtwist
– Transitioning from a non-technical background into the data world
The Alldus podcast is brought to you by Alldus, the International supplying your expert AI and digital transformation staffing needs across the U.S and Europe today. You are listening to our AI in Action series we’re leading minds and AI from across the world share their story of success and advice. AI in action cuts through the hype and explores the true impact of artificial intelligence in our world today.
JV: You’re listening to AI in Action. I’m your host, JP Valentine. Our guest today is Audrey Smith. Audrey is the CEO at MLtwist. Audrey, welcome to the show.
AS: Thank you John Paul thank you for having me. It’s a pleasure to be here today.
JV: Yeah, we’re delighted to have you, Audrey. Let’s start with yourself please. Could you give us a bit of an overview of your background in technology from where you got started, some of the roles you’ve held along the way and take us up to today as the CEO of MLtwist?
AS: I am actually a non-technical person who joined the AI industry eight years ago when I moved to Silicon Valley. I’m French as you can hear. I was actually an in-house lawyer in France then moved to the UK worked in compliance so nothing to do with AI and then when I moved to California eight years ago. I really started looking into AI because I had always been passionate about it and I learned about how it works on the web data preparation side of things, how do you create the labeling on the data that’s going to be injected in the algorithm for a lot of different projects. I got lucky enough to work on projects at Google, a lot of different projects from user experience to GDPR compliance, then I moved to Amazon where I stayed for a while working again on a lot of different projects. That has been a great school to me. I learned so much over there working on a lot of different data formats from audio, video image and text. Then I moved to LabelBox where I headed the data labeling operation team for three years and then I moved to MLtwist a year ago basically doing the same thing, helping customers with their data labeling projects and making sure they get the high quality data they need for their machine learning models.
JV: Amazing. Thank you so much for the background and great to learn about your entry into the world of AI. You’ve already worked at some of the most prestigious and well-known organizations as you mentioned with Google and Amazon but you’re now the COO of MLtwist so tell us all about MLtwist as a business: who you are what you do, what the mission of the business is, and then we can talk about what you’re doing for your customers on a day-to-day basis.
AS: So MLtwist is quite young. My co-founder founded MLtwist two years ago and I joined a year ago and it happened after he came to the realization that the data preparation ecosystem was very overcrowded. There are a lot of tools out there on the market, literally a new tool every other month, coming and they’re all good in some ways but when you, an AI company as you can imagine, working on your specific project for machine learning it’s very confusing to look at all the stores not knowing which one is going to be the best one for your own use case so MLtwist is not about reinventing the wheel it’s about making sense of this data preparation ecosystem we are trying to unify it and give as much transparency as possible to AI companies so that they can make the right choice obviously when they login to the MLtwist platform they will be able to have a fully integrated sheet and automated workflow that’s going to help them select the right tool that’s the right tool but also choose the right workforce to label their data and then get the output in the format that they want to have so there is a portion of it that is also about reformatting the data and that is not super sexy. Not a lot of companies want to tackle that issue but we think it’s very important to be able to help the customers also in some way
there and we have two different options. We can be data as a service: companies can come to MLtwist and just give us the data let us know what they want to do with their model and then from there we’re gonna pick and choose the right tool and then act as a project manager to deliver the data to them the way they want it or we can be software as a service and in that case they log into the platform and they will be able to do the entire process themselves.
JV: Amazing, thank you. I want to spend some time talking about both sides there you mentioned that and SaaS. Could you walk us through a recent project or a customer journey that would demonstrate the benefits of them using your services on it that’s side but also then as similar example for SaaS it’d be great to visualize the customer’s problem and how using ML the best is helping them solve that problem.
AS: So it really depends on the mindset of the company we’re working with. Some companies just don’t want to deal with any of that. Data preparation for machine learning is very complex but also very time consuming right and that’s not what data science teams wants to work on, they want to work on the model, they want to make a model that’s going to be great and will perform well and all the data preparation side is definitely not the sexy part of the business and they can just decide and explain us what their use case is about what they want to have in terms of requirements in terms of quality in terms of budget in terms of turnaround time because they have a deadline to deliver a product on like during the year and they need to get the data out as soon as possible and so on and so on and that’s really like where the dust piece is very interesting for them because they don’t take care of any of that we’re going to take over from there and and they’re going to get the output they need for their model and they’re going to be able to test their model based on what we deliver if there are still some issues in certain areas we can go back and then again redo the entire process and make sure that the quality comes out the way they want etc etc so that’s that’s where you just don’t want to put your hands into all of those tiny steps that are like very important but also very time consuming and we have a lot of customers like that that prefer not to touch that part of the machine learning workflow and then you have companies that have actually people that are more on the data labeling operation side of things that have been hired internally and want to manage this piece of the project they want to be part of the entire journey and they want to select the right workforce they don’t want to rely on us to do that even though we’re going to do that as well but they want to be part of it they want to give us a feedback loop on the quality that we deliver so that we can definitely improve more easily and so on to give you examples we have customers in the defense industry that just don’t want to deal with all that public just tell us what they want to see at the end of the workflow and we work towards it and we give them a sample we run a pilot with them and if they’re happy with what they see we just go scale up and then give them as many data points as they need or we’re gonna have companies in the attack industry that want to be very involved and that are going to be touching base weekly on a weekly basis and making sure that we did what we deliver to them is as good as possible and even those even some of them want to talk directly to the workforce we’re going to be working with and so on so there is no best option it just depends on what the customer wants to do and how they want to go about it and we want to make sure that we’re going to deliver what they want so.
JV: Understood. I want to spend a bit of time now talking about the team that you’ve built at MLtwist. You mentioned in your introduction as a business you’re still relatively young only founded in the last few years but you’ve already managed to bring in some really talented people to work on these complex projects. Can you give us a look behind the scenes?
AS: We’ve got the current makeup of the overall team looks like the combination of various skill sets of technical and non-technical and how it all works together on a day-to-day basis. We are currently achieve more than people so there is myself and my co-founder I’m more on the data operation side of things my co-founder is has been in the data acquisition industry for the past 10 years especially in the attack world so he’s more on the business side of things and then after that we have the full stack engineers working on the platform adding features and making sure that it’s all working well and working towards being fully automated and fully integrated with a lot of different tools that are out there and then we have project managers that are working with me on helping customers with their delivering project this is actually one of the jobs that is like the most difficult to recruit for because it’s a combination of a lot of different skills we can talk about it later if you want to and then after that we have also data scientists machine learning PhD consultant that helps us also when we build some quality control models for our customers.
JV: You are listening to the Alldus podcast When you’re looking to scale your team or if you are interested in showcasing your company in a future episode, reach out today or if you’re in the market for a new role visit our website to view open positions www.alldus.com. I definitely want to talk to you about that point of the project management side because I often get asked about opportunities for people from a non-technical background, ways they can work within AI and I think you’re perfectly positioned to to give some insight there so could you describe that role the project managers your team and how it how you have to combine lots of different skill sets to to work with your customers.
AS: Yeah, definitely. I’m very passionate about that actually, that’s what I’m seeing. I’ve been thinking about since I was at Amazon because I was helping with the coaching side of things but also at LabelBox when I was building my team and the same thing is happening here at MLtwist. It’s a very interesting role because obviously it’s it’s a non-technical world but you have to understand the technicality of all the data preparation ecosystem you have to understand what machine learning teams are talking about why they have those requirements because you’re not only about executing yourself about documenting for good practices and see how you can help them get a better workflow that’s going to be making sure that their models are going to be performing well so there are three different skills that you need to have when you are in that type of job you’re going to be able you’re going to have to be a Vendor Manager because you’re going to be working with all the different workforces that will be labeling the data for your clients so you going to be able to make sure that you’re going to have a great partnership with them you want to make sure they are happy and that everything is going well on that side then you’re going to have also to be a good account manager because you’re going to work with a lot of customers and you want to make sure they feel that they are taken care of it’s uh yeah in on that stand in that sense you have to be an account manager but you have also to understand the technicality of what they are talking about because our customers are going to be data science teams or machine learning teams and they’re going to go about it with their own technical way so you need not only to be able to understand what they are talking about but being able also to translate it in a more simpler way for the workforce that’s going to be working on the project if that makes sense and the last piece is that you have to be a good project manager because you’re going to be having this project in your hands you’re going to have the data you’re going to have some guidelines you’re going to have to understand to label the data in the best way way possible so there is no degree that you can get at the moment even though I said it in the past and I’m repeating myself here I’m pretty sure that in the future there will be a degree for data preparation Ops or data labeling Ops so the only way to find good people is to look into people who have already done that type of jobs in other companies. Somehow I like you know looking into the big tech companies that have been really into AI for a good number of years now and they have been able to train a lot of good peoples that can definitely help me in my team with the customers yeah it’s such an important one because the AI field as a whole is constantly
evolving and emerging and new positions pop up all the time which haven’t previously existed so they’re very difficult to stop an oil for recently I think of the position of ml Ops well as somebody who’s involved in recruiting in your previous companies and now at MLtwist when you find candidates who have not held that title before but you want to describe the role to demand why it’s attractive and exciting.
JV: What are some of the things you tell candidates that would get them interested in a position like this?
AS: I think it what I just mentioned about which is the fact that this is a very unique position where you’re going to learn so much about different type of skills being able to talk to customers in some ways and have direct relationship with the customer and actors and account manager is already exciting in itself but you don’t get to to be only that you’re going to be also like having the whole control over what’s going on by being also the project manager managing the entire end-to-end workflow and also being the vendor manager and making sure that things are going to be going well in that side of things and from the different people I got to hired in the past five to seven years they really like that that this is not just one type of job this is a combination of three different jobs and you can get for it and one thing that’s very exciting about that position is that when you work with AI companions a lot of them you’re gonna see the application on what they are doing with machine learning in their own ways what is the use case about and I
worked in so many Industries in so many different use cases and it’s just like it you can get four because it’s just amazing to see what AI is doing into like how it’s impacting a lot of different Industries in so many different ways and you just being part of that Evolution and that’s so exciting to me.
JV: Audrey I want to spend a bit of time now talking about what’s next for MLtwist what the project roadmap looks like for the next 12 to 24 months what you guys are working towards where you’re trying to grow the company could you give us an oversight of what the next 12 to 24 months are going to look like both from a customer acquisition perspective but also the growth of the team and what opportunities there are going to be for people to come and join MLtwist.
AS: MLtwist Focus right now is to integrate with way more Partners we have already quite a handful of them but we want to make sure that we touch on all the different tools out there from
synthetic data, data augmentation but all the different labeling tools out there that that can help in a lot of different industries so the idea is to expand our Marketplace and our Partnerships with all those tools but also try to accelerate our customer acquisition rate because now we feel ready to go after also Enterprise companies and offering them services that they don’t have at the moment basically if I want to make it short a lot of Enterprise companies are using already a lot of different tools out there but they are not integrative and so it’s very complicated to communicate from a team to another one understand what everyone is doing about is watching on and also leveraging the knowledge of a team for another one that’s something that we really want to go into helping Enterprise companies get more transparency than making transparency sorry and making sense of what they have already built internally by using their Matrix platform.
JV: Final question for me then Audrey as someone who successfully made the transition from non-technical into now a very technical company overseeing both so it’s speaking to an audience of data professionals or people who are interested in the space What are some of the key things you’ve learned successfully making this transition and for anyone interested in making the transition what can they do to improve their chances
AS: I think it’s about going for it I would have never thought I would end in that type of position myself when I was in France or even when I was in the UK for sure but I just went for it when I arrived in Silicon Valley the energy was so strong and the opportunities were available and so I decided to give it a shot even though you know like the first few months at Amazon I was brought on like very technical meetings and I didn’t know what they were talking about. I was very conscious about what was going on and then I just decided to Deep dive and believe in myself and I kept going and it was like obviously very transparent to me what was going on and what I needed to do so the idea I think especially for women is to not be intimidated by the fact that you’re not a technical person. I think it’s really about going after what you want and you’re gonna get there eventually it’s just a question of learning on the side and but also asking questions over and over. My boss at Amazon was a PhD in computer vision and he taught me everything I needed to know on the data labeling operation side of things so it’s just about making sure that you talk and you raise your voice when you have a question and you just keep at it and it’s going to happen especially in that type of position where you just need to understand the technical interesting and you can use your own skills your business skills your analytical skills to bring more to the table and be a different asset to a machine learning team.
JV: Audrey thank you so much for coming on and talking to us today great to learn about your background amazing to hear about what you’re doing at MLtwist and creating almost a new profession within the world of AI it’s very exciting we wish you the team and everyone at MLtwist the best of luck in the months and years to come and we look forward to having you back on the show in a year or two and hearing how everything has progressed from there
AS: Thank you so much that’s great thank you so much.