Today I’m pleased to announce that we raised $9.5M in new funding led by Sequoia Capital. Bill Coughran, partner at Sequoia, will join our board. Our existing investors‒Redpoint Ventures, Zetta Venture Partners, and XSeed Capital‒all participated in the round.
Series A funding indicates two milestones in an enterprise company’s life: strong revenue and momentum, and a compelling plan for the deployment of new capital. It also marks the start of a new partnership, in this case among us, Bill, and Sequoia. We are thrilled because Bill is that unique leader who has contributed to science, managed large technical teams, and led businesses. Sequoia has also funded the businesses we admire most, among them Google, Apple, and Stripe. In this post, I’ll describe what we’ve achieved, what we plan to do, and why we are certain that Bill is the right partner to help us do it.
How We Got Here
John and I started Lilt because we believe that a person’s native language shouldn’t limit their ability to learn, grow, and support themselves. Were I to have been born speaking Malagasy, and I had to learn English to study computer science, then I would have been disadvantaged from the start. This is an overlooked form of social inequity: you cannot control your native language. Social problems rarely have satisfactory technical solutions. However, steady advances in computational linguistics, our field, have brought us to a time in which universal information access no longer seems like a fantasy.
Our contribution is to enable businesses to offer the same experience to all customers and employees irrespective of language. Revolutionary consumer tools like Google Translate have empowered people to pull translated information instantly and on-demand. There is no quality guarantee, but that is an acceptable compromise given the speed and convenience benefits. Now we want to give businesses these benefits yet with quality guarantees. This guarantee is not free, but the degree to which we can make it affordable is directly correlated with the degree to which a business can democratize its information.
In the fall of 2017 we began offering a complete, high-quality, and affordable solution for enterprise translation. Our customers have worked wonders. For example:
- Zendesk now uses a human+machine solution to translate all of their very large knowledge base into 10 languages (up from six).
- Canva is translating 100% of their Smart Templates‒their core asset‒into more than 20 languages.
- A major financial media company has reduced the time-to-publish on their terminal from two days to 50 minutes.
There are others, including governments, who will be announced in the coming weeks and months. Each business is unique, but each of our brave early adopters believes on principle that information should be universally accessible. They are willing to embrace radically new approaches to translation to achieve that goal.
We succeed when our customers succeed. Our revenue has doubled in each of the last three quarters. In June, our revenue was twice all of 2017, and our translators produced 1.2M words of high-quality translation. We’ve grown from six employees to 21, released two neural MT systems, and published an exciting new paper on adaptive learning. We work with more than 300 highly skilled domain specialists, which represent less than 5% of the people who have applied to work with us. We’ve expanded our offices in both San Francisco and Berlin.
Where We’re Going
Enterprise localization is a complex production process. Our choice to integrate vertically means that we must internalize the complexity of the whole process: from the development of new algorithms to sophisticated customer support and workflow integration. This requires operational discipline, a service disposition toward our customers, and coordinated investments. The first two requirements will be the responsibility of the people we hire and the culture we build. The investments in people, data, algorithms, and workflow are enabled by both the growth of our business and by fresh funding.
We built Lilt for translators, and we seek to offer the best work environment for them. This means:
- Recurring work and an associated feeling of recognition and ownership
- Fast and direct customer interaction
- A world-class, human-centric translation workflow
- 48-hour payment terms
Our translators directly inform our product choices. If our product helps translators do great work, then our business customers derive greater value.
At Lilt, translators will never be asked to post-edit. We believe that the MT post-editing / rate reduction business model leads to the sweatshopification of language work. That’s bad for the world. If you’re great at your craft, then please join us.
We also built Lilt as a place where our employees can do the best work of their careers. We’re hiring more people, but we’re also investing heavily in the personal growth of each member of our current team. All of us care deeply about language, and the pairing of that mission with a deeply technical culture is what leads to innovation. Join us if you care about language, too.
Machine translation research takes place in the open, so differences between systems are almost entirely explained by data advantages. Because our technology is adaptive, it can specialize as it is used. This yields a unique dataset for developing not only machine translation systems, but also more general workflow tools. For example, we are investing heavily in multilingual error correction tools to automate quality control.
We also need to accumulate more general domain training data to serve (a) businesses who are new to translation, and (b) languages for which few data exist. We will invest heavily in both areas.
Our core research team has published over 130 papers, and has 70+ years of experience in MT research. Research is the heart of our company. Bill is a unique partner in that regard, having completed a PhD in computer science at Stanford under Gene Golub. He then led Bell Labs, and then engineering at Google before joining Sequoia in 2011.
Now is the most exciting time in the 70-year history of MT research. We believe that human+machine systems are best built holistically, so we will continue to develop our decoder and associated NLP components. If you are a world-class researcher who also likes to build, and you want both creative freedom and the dynamism of a startup, then we want to talk to you. Join us.
Localization is a horizontal business function: businesses of all sizes and in all markets do it. Every workflow is different, not because translation work differs fundamentally across businesses, but because organizations are fundamentally different. Unique collections of people, processes, and systems give rise to unique requirements. Some organizations outsource while others have partial or complete in-house teams. Some organizations are stuck with legacy or purpose-built information systems while others use new SaaS systems. Some are led by former translators and project managers while others are led by people whose interest in language is strictly operational. Some IT departments permit hosted solutions while other demand on-prem.
To that end, we have built a component-based, API-first product that can be adapted to a broad range of business settings. For example, Sprinklr uses our web application for translation and project management. They use a mixture of in-house teams (for two languages), internally managed freelancers, and Lilt-managed professionals. They use both manual and API-level content flows. Crucially, this is all achieved through one product with one, unified pricing model for easy budgeting.
Over the next 12 months, we’ll invest heavily in the following:
- Data security and enterprise readiness ‒ some of the most secure businesses and government agencies already trust us, and there is still more that we can do to protect our customers.
- Reporting and supply chain visibility
- File conversion and content connectors
Of course, we’ll continue to build out our core translation workflow (in collaboration with our translators) and our API.
Our New Partnership
Venture capitalists invest in venture-backable businesses. Because it is riskier to invest in a startup than, say, Starbucks, the expectation of a reward is much higher. Successful startups must have some uniquely defensible competitive advantage, and they must address a market large enough to generate outsized returns. Some markets are appropriate for venture capital and others are not. Enterprise software and digital advertising are two markets with successful venture-backed companies. Landscaping and commercial fishing, while large markets, are not (although modern venture capital has its origins in 19th century whaling).
The jury is still out on localization. It is a large yet highly fragmented market. As currently organized, it has no inherent economies of scale, and thus has never produced either a billion-dollar business or a good venture-backed exit. It has been slow to embrace technology, and has thus ceded development of core language technology to companies such as Google and Microsoft that have no intrinsic interest in the market. It is often characterized by cozy vendor/buyer relationships that stifle innovation and progress. This is disappointing, for the localization problem is so vital, and thus rapid progress would obviously benefit humanity.
Bill is among the vanguard of investors who believe that the impact of AI will be most profound in the enterprise. Tools are being built that allow 20 people to do the work of 100. This is the augmentation view of AI first advocated by Doug Engelbart in the 1960s, and which is now becoming a reality. Specifically, machine translation is having its moment after 70 years of work by pioneering figures in computer science. Bill believes that Lilt is uniquely positioned to help businesses operationalize those advances, and that more businesses than ever can be equipped to go into more markets than ever.
This is a unique time in the history of both commerce and technology. True, success results in financial reward, but this is one of those unique cases in which the benefit to humanity is incommensurately greater.
This started for me in both a literal and figurative desert in 2005. Over the past 13 years I have used, researched, built, and commercialized language systems. Only in the last few months has Lilt began to startle me. I still translate, and just yesterday while working there were several moments in which the system proposed things that I never would have thought of.
For many years I was like the proud parent who gave his child the benefit of the doubt. The shortcomings were obvious, but the growth and progress were significant. Now not only the technology but also the business is emerging as its own entity with capabilities separate from and unexpected by me.
I thank our team, investors, and customers. I thank Bill and his team for offering their world-class experience in company-building. And I thank John for the most fruitful collaboration of my life.
We enable our customers to serve theirs. Then their customers have knowledge, they are empowered, and when they are empowered, progress is made.
Let’s make progress together.