Technologies
Yandex technologies create new possibilities and help shape the future for millions of people.
SearchSelf-drivingNatural Language Processing
Computer VisionVoice ProcessingNowcasting
We are driving innovation in AI and its everyday use. Having pioneered information retrieval in the early 90s, today our core AI technologies power computer vision, speech recognition, speech generation, machine translation, self-driving, among many other Yandex solutions, products and services.
Search
Established in 1997, Yandex Search provides contextually relevant information to millions of users on multiple platforms for both text and voice queries. As one of the world’s largest search engines, Yandex Search accesses a global web index crawling billions of documents, including text, images, and video to provide personally relevant and time and place specific results for every individual user. We integrated gradient boosting on decision trees in search with the introduction of our Matrixnet algorithm in 2009, and have been continually advancing the quality and versatility of our search technology since then by introducing new intelligent updates and other state-of-the-art innovations.

Self-driving technology
As of early 2021, almost 200 autonomous cars and delivery robots powered by Yandex’s technologies, infrastructure and engineering talent, drive on the roads in Russia, Israel and the US, with over 6 million autonomous miles behind them. Having started developing self-driving technology in 2017, Yandex launched Europe’s first robo-taxi service in 2018 and Russia’s first autonomous delivery robot Yandex.Rover in 2019. Our self-driving vehicles combine machine learning, navigation, mapping, and cloud technologies, which together with LIDAR sensors, cameras, and radars provide our cars with a capacity to operate in diverse driving environments and weather conditions.

Automated translation
Our machine learning technologies power Yandex.Translate, a public machine translation service that supports translation between nearly 100 languages, including those that are rare or dying out. Leveraging Yandex’s huge index of web content and distributed machine learning infrastructure, Yandex.Translate provides world-leading translation quality for text, and large documents, websites, audio, and image files.

Content-based image retrieval
Our image retrieval technology can access a massive index to search images based on their content, such as colors, shapes, or textures. The reverse image search and image recognition technologies that we develop use convolution neural networks, joint text and image semantic embeddings, face features, local image descriptors, and other techniques to retrieve images similar or related to the sample image, including altered or manipulated versions.
Super-resolution imaging
DeepHD, our super-resolution technology based on neural-networks, enhances the quality of both images and videos using Generative Adversarial Networks (GANs). It boosts resolution and restores details, removes noise defects, and compression artifacts. DeepHP was the world’s first technology of its kind to be applied to images and videos hosted on popular video and live TV streaming services with multimillion user audiences.

Optical character recognition
Yandex’s OCR technology locates and recognizes text on images. Our OCR yields especially advanced detection of text “in the wild” or text captured by mobile cameras in unconstrained environments and scenes such as signs, packages, and menus. OCR also helps to deliver instant image translations for Yandex.Translate.

Yandex's automatic speech recognition has been critical to making our virtual assistant, Alice, the most popular spoken voice assistant in Russia. Built on end-to-end neural networks and enhanced with spectral augmentation techniques, our ASR stack has been developed to provide a reliable and user-friendly speech interface for our users. Working together with device manufactures and machine learning engineers, our technology solves tasks from hot-word detection and voice quality enhancement, to speedy delivery of partial results and robust end of utterance detection.
Our in-house text-to-speech technology has played a vital role in the creation of the character of Alice, allowing us to communicate with our users in her own unique voice. Current projects include experimenting with fine-grained style control, a technique that allows us to project specific contours and subtle prosodic effects onto the voice of Alice. Text-to-speech is also a critical component in the accessibility features for Yandex products and services.
Nowcasting
Our weather forecasting technology Meteum blends five traditional weather forecasting models with machine learning using Yandex’s open-source gradient boosting library, CatBoost. Meteum also integrates nowcasting, or real-time precipitation forecasts within hyper-local areas, as small as a single block of a neighborhood, anywhere in the world.