Yandex is holding Weekend Offer ML — an online hiring event where the full interview cycle is compressed into two days: on September 12–13, 2026, candidates complete two technical sections, final interviews with teams, and may receive an offer. Applications are accepted until September 4; the event is looking for ML and DL engineers with at least two years of experience in NLP, CV, RecSys, and Classic ML, as well as product and data analysts.
What happened
The event is held entirely online. After an HR screening, candidates solve two technical sections on September 12 and meet for final interviews with teams on September 13; at the end of the weekend, successful participants may receive an offer. The team lineup includes R&D Alice AI, working with LLMs, VLMs, and ART image generation; “Alice and Smart Devices” with a focus on tool-calling, SFT, and RL; “Search with Alice AI,” developing agentic search on the Nova platform; “Independent Ecom” with ranking of 2 billion product offers; and “Advertising Technologies” with Europe’s largest recommendation system.
Context
Fast hiring has long ceased to be a one-off event for Yandex: the company has an official calendar of hiring events, where an offer in one to two days is presented as a regular format. This allows the employer to shorten multi-week interview cycles and compete for ML talent with a shorter funnel. The list of teams also serves as a public snapshot of the production stack: multimodal VLMs and image generation, tool-calling, SFT, and RL in production, agentic search on the Nova platform, and RecSys at the scale of billions of offers. According to Yandex, about half of search traffic already comes from agentic scenarios, and the company is concentrating hiring around agentic platforms.
Why this matters for the industry
For the industry, the team lineup is an open map of Yandex’s investments in ML and a guide to in-demand skills: the company directly shows where it has heavy production tasks — agentic search on the Nova platform, tool-calling and RL for Alice, VLMs and image generation, and ranking of 2 billion product offers. The shift in demand from pure SFT fine-tuning to agentic training and evals for tool-use suggests to builders which directions products and integrations will be built around. If the two-day format shows sufficient conversion, other large employers will likely start copying it, and pressure to shorten hiring cycles in the Russian market will increase.
Why this matters for users
For a reader with two or more years of experience in NLP, CV, RecSys, or Classic ML, this is a chance to receive an offer in one weekend without quitting their current job: everything is online, and only a laptop is needed. The window for action is short: applications must be submitted by September 4, and the sections will take place on September 12–13. According to the organizers, interview results will be saved for future applications, so participation makes sense even for those not yet ready to move to a new job. Preparation can be done using the official guide to ML interviews on yandex.ru/jobs, medium-level tasks on LeetCode, and in Yandex Contest.
What is still unknown / limitations
The two-day format is a compromise in evaluation quality: a small sample of interviews increases the variance of decisions compared to a multi-week cycle, and public preparation materials only partially reduce the noise. The share of agentic scenarios in search traffic stated by Yandex remains a vendor estimate: public materials do not define an “agentic scenario,” the measurement period, or the methodology, so the figure should be considered uncalibrated until independent verification. The preservation of interview results for future applications is the word of the organizers; participation terms should be clarified on the event page.
Sources
Author
Look at AI, editorial team