Data Science — about agents and LLMs

We already have agents that work in support, communicate with users, solve some things themselves, and hand off to a human in others. However, there is much to improve in retrieval, add work with attachments and tools, build an evaluation system around it (LLM-as-Judge, dialogue simulations, datasets), and think about memory and architecture. If you build quality evaluation first and then the product, love OpenAI Agents SDK, and generally strive to make an agent behave predictably — we are on the same page.

Honestly, we are not just looking for specialists, but for those who get excited about such tasks themselves. We have a cool team, a lot of freedom, and a desire to build an AI direction that we can be proud of.