RocketHuntSmart job search
Reach out directly about this role
By city
3-5 years
Experience
Full-time
Employment
Hybrid, Onsite
Work Format
Middle
Grade
Backend
Specialization
AdTech
Industry
Corporation
Company Type
Developer for the Automatic Ad Generation Team
Automatic ad generation is the infrastructure that turns products on the internet into advertising creatives, taking into account advertisers' settings. Our platform is built on the BigRT streaming processing framework on top of YTsaurus: thousands of machines in our service process changes for over 10 billion products / services / website catalogs from across the Runet with second-level latencies.
For generation, we use not only algorithms but also YandexGPT — for generating ad headlines and texts, various neural networks — for ranking ads and assessing the appeal of generated banners. We also participate in developing the neural network inference infrastructure and have managed to integrate processing with calls to the model service.
Optimizing real-time processing infrastructure Your tasks will include increasing throughput, CPU profiling, and saving thousands of cores. You will also make the service recoverable within minutes during downtime.
Developing ML generation infrastructure You will implement new models, optimize the incoming flow to the service, and conduct regular drills.
Improving quality We want ads with high conversion potential from all Runet products to get into impressions — so advertisers get more purchases, and the ad engine doesn't waste resources on storing, ranking, and selecting low-value banners. For this, we have a service that makes on-the-fly decisions about how many ads a specific advertiser needs, ranks the generation results, and selects winners.
Influence on the product One of our tasks is to understand what advertisers lack and which factors influence how well a product will sell. To solve this, we implement new ad types and expand the set of available elements for banner generation.