AiThority Interview with João Graça, Co-founder and CTO at Unbabel
Please tell us a little bit about your journey and what inspired you to start at Unbabel.
I am co-founder of Unbabel and currently serve as Chief Technology Officer. Our vision for Unbabel is to become the “World’s Translation Layer” by eliminating language barriers between companies and their customers through a combination of bespoke Neural Machine Translation (NMT), advanced automated quality estimation, and a global human network of tens of thousands of ‘Unbabelers’, who post-edit the automatic translations to professional quality.
The other founders and I saw an opportunity when we noticed a glitch in the way the world was consuming information – the internet provides us content from all over the world however, we could not properly read the contents when relying solely on machine translation.
What is a LangOps platform and how has this technology evolved in the last 3 years?
At its core, LangOps (language operations) is a holistic approach to multilingual communication. The technology gives organizations transparency and control over every aspect of language across their business. Recently, we have seen the demand for tech such as LangOps drastically increase as a result of a growing globalized market. In fact, 65% of customers prefer content in their own language. This has meant that LangOps has had to rapidly evolve to meet this demand, and that has seen many advancements as well as a growth in industry recognition – last year, Unbabel was recognized as a sample vendor in 2022 Gartner Hype Cycle for Natural Language Technologies.
With the rise of AI and machine learning capabilities, what is the future of retail technology platforms?
A great example of the rise of AI and ML capabilities when it comes to retail technology platforms is the growth of the recommerce market or the reselling of goods through popular websites such as Poshmark and eBay. In fact, the recommerce market grew twice as fast as the wider retail market in 2021, and is expected to reach $289 billion by 2027. That said, with this recent boom in activity in the recommerce world, the need for translated recommerce content is at an all-time high to meet market demand.
While many brands use generic translation services for their recommerce translation needs, these platforms simply can’t keep up with speed, scale and demand the growing market demands. Enter AI and ML in end-to-end translations. With the addition of AI, specifically human-refined AI as seen in Unbabel’s platform, machine translation is able to stay on par with the speed and scale needed. Beyond simply adopting an AI-powered platform, the Unbabel platform features a global community of skilled human editors who fine-tune translations and send those refinements back to the ML algorithm in an effort to continuously improve.
The bottom line is, the retail market, recommerce included, is showing no signs of slowing down and its direction is worldwide. As we move into an increasingly ecommerce centric world, translation services need to be part of the equation.
How did the pandemic change the AI and ML landscape?
As we move into an increasingly remote world, the need for AI and ML solutions across the board has increased, including LangOps. From a customer service perspective, support via agents comes with challenges that neither localization nor machine-only translation could handle.
For example, at the peak of the pandemic, the travel industry was inundated with cancellations around the world and without AI / ML translation support, the demand, scale, and quality of translations needed to solve these issues simply could not be met. At the same time these cancellations were happening, many businesses were also scaling down their teams and the option to outsource support was not a scalable solution for dealing with customer requests in different languages. However, with the addition of AI-powered, human-in-the-loop ML, multilingual communication issues can be more easily solved and more importantly, easily scaled to meet demand as it inevitably increases.
What is Unbabel and what are your core offerings? What is your vision for the company and the industry in general?
At Unbabel, we empower enterprises that are seeking to conquer new markets to deliver multilingual customer support at scale by breaking down language barriers with our LangOps software. Our bread and butter is The Unbabel LangOps platform which combines leading MT technology with humans-in-the-loop to produce high-quality multilingual translations. Essentially, we combine the speed and scalability of machine translations with human translators to create the perfect combination. We then utilize feedback from our human translators to retrain the AI so that it evolves, grows smarter, more fluent, and over a period of time, needs less human input and correction. Our new offering, Unbabel Self-Service, now allows anyone to benefit from trialing our award-winning LangOps platform. And with Quality Intelligence, Unbabel can provide full transparency into the quality of translations, to help prevent errors before they occur.
We started with an API and a vision to be the world’s translation layer, but in recent years with the shift to work from home, we have taken our vision a step further to truly break down language barriers and make this shift simpler for businesses and individuals alike.. As opposed to the siloed approach we often see, where sales uses native speaker translators and marketing uses a LangOps Platform and customer service uses something else entirely, we are striving to make one cohesive system and break down our language barriers. Ultimately, we want to see AI-powered, human-refined machine translation technology serve as the centerpiece to power an entire organization’s language strategy and make language a seamless experience for all.
Tell us more about Zendesk-Unbabel integration.
When a business utilizes Unbabel’s Zendesk integration, it enables Zendesk agents to deliver a multilingual customer experience thereby allowing a business to more easily scale and meet this growing demand for translations. Essentially, Unbabel acts as a force multiplier for a business’ Zendesk team, turning individual agents into an instant army of translators able to address any customer anywhere in the world — without hiring headcount for language specialist support. With this comprehensive and seamless integration, businesses can reduce cost, meet surges, and support growth to deliver exceptional customer service worldwide with no disruptions to agents’ workflow.
Any other integration in Conversational AI or Voice market that you are particularly proud to talk about?
We’ve been working on several areas in Conversational AI to strengthen our CS solution, namely, intent classification to make the dialog more effective, sentiment classification to understand how to better address the end customer and conversation quality estimation that aims to quantify how good the entire thread is. The outcomes of these research products should be incorporated into our products during 2023. We’re also continuing our research agenda on large language models. After the success we had on using this sort of model for quality estimation, we want to apply them to other areas such as our machine translation.
What’s missing from LangOps that you would like to plug in the next 2 years?
We are starting to add quality evaluation capabilities so that LangOps can see the quality issues in every single translation and proactively act upon it. The next step is to make the customer data lake, all data from the customer that is visible and actionable – where you can see an error and trace that error to the original place it was generated, where it was translated, and fix across your data lake in one go. After that we will add more language capabilities to the platform that will leverage the customer data lake, such as named entity recognition, sentiment analysis, and content creation. This is when the full view of Language Operations will be accomplished.
Please tell us a little bit about your IT stack and how it has evolved in the last 2 years to support your operations?
We are based on the cloud under kubernetes, the biggest change over the years has been the increased need to deal with larger models and use GPUs. On top of that we are a python and go shop, python for most ML parts and Go for the more transactional parts. We build most of our stack internally.
What kind of benefits do customers of an AI-powered LangOps company derive from their investments?
Business translation services are a must-have for organizations looking to expand beyond their home country. Even beyond global expansion, business translation opens up a slew of opportunities that may inspire key stakeholders to pursue globalization including a larger customer base, more far-reaching talent pool, and increased revenue potential.
That said, relying 100% on fully-automated AI language solutions is fast and cost-effective, but the lack of human input often results in subpar quality that leads to bad customer service experiences. In Unbabel’s 2021 Global Multilingual CX Report, 92% of respondents said that poor quality customer support in their native language will negatively impact their trust and loyalty toward a brand a moderate amount to a lot yet the opposite can help a brand to rapidly expand and create a happier customer.
In fact, Unbabel customer, ClassPass, was looking to expand from the United States into international markets, but the daunting task of having customer service in every language felt overwhelming. Yet with the adoption of Unbabel, ClassPass was able to scale its support from one language to 15 languages, allowing the company to grow its subscriber count and high-quality customer experiences.
As yet another example of how the pandemic changed AI-powered LangOps, another Unbabel customer, Logitech, experienced a 300% surge in tickets after the work from home shift related to its home office products. As a result, Unbabel powered their translation across 24 different languages, helping Logitech close tickets within 48 hours instead of 13-15 days.
Using human translators to address the quality problem is effective; however, taking a human-only approach to business translation is often costly, time-consuming, and difficult to scale and this is where our LangOps approach comes into play.
What is your take on advancing technologies in Edge computing, AIOps and IT security?
AI Ops is a fundamental part of our strategy and something we’ve been developing deeply, it is the base for the LangOps platform.
Please tell us about the major milestones you manage to achieve in 2022.
2022 and into early 2023 was a wonderful period of growth for Unbabel between an acquisition, multiple new appointments in our company and of course, technological expansion and innovation. Most notably, we were proud to announce a new Zendesk messaging integration which allowed real-time translations via WhatsApp as well as a partnership with Ada which provided multilingual assistance to live agents and even more language capabilities. Both of these partnerships and product initiatives contributed to our ultimate goal of the “World’s Translation Layer.”
In terms of new executives, we welcomed our new Chief Revenue Officer, Jennifer Bers and new Chief Product Officer, Alex Cobb who took the helm to continue driving our vision of breaking down language barriers at Unbabel. Last but certainly not least, we kicked off 2023 with a bang with our acquisition of EVS Translations, a single-source translation supplier for all language requirements and market leader for premium human translations. With this acquisition, we were able to expand our existing capabilities and continue to innovate to deliver the most comprehensive multilingual customer experiences for our customers.
What AI initiatives are you currently focusing on and why?
We are particularly excited about our new Unbabel Self-Service and Quality Intelligence offerings as they will serve to offer greater accessibility to LangOps and break down more language barriers which is always our goal.
The free 14-day trial of Self-Service we are offering gives anyone access to Unbabel’s full capabilities, and enables them to quickly and easily set up translation pipelines using AI combined with human editors, request translations, and integrate with Zendesk. Alongside the Self-Service offering, we announced free public access to our Quality Intelligence solution enabling anyone to receive a demo using Unbabel’s industry-leading quality estimation technology.
With the introduction of both products, both businesses and the general public can benefit from features like the ability to set up translation requests that utilize a combination of Unbabel’s AI-powered LangOps solutions and team of translators, access to our quality estimation AI-technology developed using reams of human input data to predict the quality of every translation, and more. Simply put, we are passionate about what we do at Unbabel and with these developments, we are making our technology more accessible and further contributing to the goal of breaking down language barriers in CX.
Thank you, João! That was fun and we hope to see you back on AiThority.com soon.
[To share your insights with us, please write to firstname.lastname@example.org]
João Graça is a co-founder, Chief Technology Officer, and computational genius behind Unbabel. Portuguese born, João studied computer science at doctorate level at one of Lisbon’s most well-respected technical universities, Instituto Superior Técnico de Lisboa. During his studies, he published a number of well-received papers on machine learning, computational research, and computational linguistics — all of which form the bedrock of Unbabel’s machine translation engine. After graduation, João worked with INESC-ID, developing research in natural language processing (NLP) and went on to do his postdoc in NLP at the University of Pennsylvania. João was awarded a Marie Curie, Welcome II Scholarship (2011), which he declined in favor of entrepreneurship. He worked with now Unbabel CEO, Vasco Pedro, together on the development of language learning algorithms and machine learning tools, plus held various research scientist roles before co-founding Unbabel in 2013.
Unbabel empowers enterprises that are seeking to conquer new markets to deliver multilingual customer support at scale. The company’s award-winning Language Operations (LangOps) platform blends advanced artificial intelligence with human editors for fast, cost-effective, high-quality translations that get smarter over time. Unbabel integrates seamlessly into existing workflows within all major CRM systems, including Zendesk, Salesforce, Freshworks, and Intercom, to make your customer service agents instantly multilingual. This enables businesses to provide exceptional experiences to customers in their native language, anywhere they are in the world.
With headquarters in Lisbon, Portugal, and hubs spanning the globe, Unbabel facilitates multilingual customer service interactions in more than 130 language pairs for leading brands such as Facebook, Microsoft, Booking.com, Panasonic, Logitech, and more.