Lingo.dev is an application localization engine for developers

Unprecedented one of the unilaterals who want to communicate with the international masses. Translate Old Old Translate can convert the entire photo and sound content and websites across hundreds of languages, while modern tools such as ChatGPT also act as translators in a useful pocket.

At the end of the background, Reckless and Eleven It has reached high -end billion dollars assessments of the various intelligence related to the language, which companies can convert into their own applications. But the new player is now entering the battle, with a detective localization engine that serves the infrastructure to help developers go to the world-a “tape” to localize the application, if you will.

Previously known as Replexica, Lingo.dev The developers who want to make the front end of their application are entirely translated from the beginning; All they need to worry about it is to charge their code as usual, with Lingo.dev coming out under the hood on the automated pilot. The result is that there is no text.

Today, Lingo.dev calculates customers like French unicorn is bad artificial intelligence and Cal.com competitor. To lead the next stage of growth, the company announced that it had raised $ 4.2 million in a seed of funding for the leadership of the capital, with the participation of Y Combinator and a large number of angels.

I found in translation

Lingo.dev is the manual work of the CEO Max Brillutsky And CPO Veronica Brillotskaya (In the picture above) who announced that they sold a precedent saas company called Concept to The buyer did not reveal it last year. The duo has already been working on Lingo.Dev since 2023, with the development of the first initial model as part of a Hackathon at Cornell University. This led to the first paid agents, before joining the YMBINATOR (YC) fall program last year.

In essence, Lingo-Dev is an interface programming interface that can be called developers locally Through Cli (Order of command line), or through direct integration with the CI/CD system via GitHub or GitLab. Therefore, in essence, development teams receive withdrawal requests with automatic translation updates whenever a standard symbol is changed.

At the heart of all this, as you might expect, it is a large language model (LLM) – or several LLMS, to be accurate, with Lingo.Dev format between the different inputs and outputs between them. This combination and matching approach, which combines models from anthropology, Openai, is among other service providers, to ensure the choice of the best task model.

“Various claims work better in some models on other models,” Prilutskiy explained to Techcrunch. “Depending on the state of use, we may want a better time, or the cumin may not matter.”

Of course, it is impossible to talk about LLMS without also talking about the privacy of data – one of the reasons why some companies It was slower To adopt artificial intelligence. But with Linggo .

“We don’t expect to send any personal data to us,” said Brillutsky.

Through Lingo.Dev, companies can build translation memories (pre -translated content store) and download their style manual to customize the brand voice for different markets.

Companies can also determine rules on how to process certain phrases and in cases. Moreover, the engine can analyze a specific text mode, and make necessary adjustments along the way – for example, a word when translated from English to German may double the number of characters, which means that it would break the user interface. Users can direct the engine to circumvent this problem by reformulating a piece of text so that it matches the length of the original text.

Without the broader context of what is already what the application is, it may be difficult to localize a small part of the independent text, such as the sticker on the interface. Lingo.Dev wanders this using a feature called “consciousness of context”, as it analyzes the full content of the localization file, including the keys to the adjacent text or the events system that sometimes contains translation files. It is all about understanding “MicroCONTEX”, as well as Prilutskiy.

More than that comes on this front in the future, too.

He said: “We are already working on a new feature that uses the application’s user interface screen shots, which Lingo.dev will use to extract more contextual hints about the user interface elements and its intention.”

Local going

It is still somewhat early for Lingo.dev in terms of its way to full localization. For example, colors and symbols may have different meanings between different cultures, which does not meet Lingo.dev directly. Moreover, things such as metric/empire transfers are still something to be addressed by the developer at the code level.

However, Lingo.dev supports Messageformat The framework, which undertakes the differences in pluralism and special formulation between the sexes between languages. The company has recently released an experimental experimental feature specifically for expressions; For example, “to kill two birds with one stone” that has an equivalent in German, translated almost into “to hit two flies with one swat.”

Moreover, LingoDev also performs AI’s research to improve different aspects of the automation process.

“One of the complex tasks that we are currently working on is to preserve female/masculine versions of names and verbs when translating between languages,” said Brillutsky. “Various languages ​​codify different amounts of information. For example, the word” teacher “in English is neutral, but in Spanish either”Maestro(Male) orTeacher(Female). Ensure that these nuances are kept properly decreased in light of our applied research efforts. “

Ultimately, the game plan is more than just a simple translation: it wants to approach what you can get with a team of professional translators.

“In general, (the target) with Lingo.Dev is to eliminate friction of localization comprehensively, so that it becomes a brown infrastructure and a natural part of the technology group,” said Brailotsky. “Similar to how the tape canceled frictions of online payments to the point that it has become a basic developed tool group for payments.”

While the founders were recently in Barcelona, ​​they were transporting their official home to San Francisco. The company calculates only three employees, as it is the engineer’s founder – and this is a lean starting philosophy that they plan to follow.

“People in YC, I and other founders, we are all huge believers in that,” said Bilutsky.

The previous start starting, which provided the concept analyzes, was completely obtained with high-level customers including Square, Shopify, and Sequoia Capital-and had a total large number of zero employees outside MAX and Veronica.

“We were two people, full time, but with some contractors for different things from time to time,” Brailotsky added. But we know how to build things with the least resources. Since the previous company was equipped, so we had to find a way to work. We repeat the same lean style – but now with funding. “