Wrote your first agent in LangGraph and want to level up? We in the Alice team are building a multi-agent platform that will solve various problems for millions of users — from personal productivity to commercial scenarios. WebAgents, tool calling, (V)LLM, A/GRPO, simulators, hardcore R&D — this is all about us. We've already figured out the GPUs, so we're looking for cool colleagues to build successful mass-market agents with.

What tasks await you

Development and implementation of LLM-based agents, design of ML components for the agent platform You will create agents that can handle complex, multi-turn, and multi-modal user requests. You will develop task routing systems, tool and model selection, as well as configure function call management and orchestration of various models within the pipeline.

Replacement of traditional classifiers with LLM pipelines and RAG systems You will need to transform old architectures into modern pipelines with LLM, function calling, planning, and reasoning.

Fine-tuning, configuration, and integration of models, inference optimization You will work with both open-source and internal models, including fine-tuning, RL, and adaptation for the agent platform. You need to ensure fast and stable agent performance in production, including inference optimization and interaction with infrastructure (Eagle, FP8, distillation, multi-dimensional parallelism, etc.).

Research and experiments with agent graphs and memory We experiment with agent calls, memory subsystems, behavior simulation, reasoning architectures, and planning systems.

Participation in the development of RL models for training dialogue agents If you have experience with or interest in multi-turn reinforcement learning, you can dive into setting up and implementing such experiments at the scale of Yandex and Alice with hundreds of tools and many turns.

We expect you to

• Have deep knowledge in Deep Learning and NLP
• Have experience with LLM and pipelines around them (function calling, memory, RAG)
• Have implemented ML systems in production
• Think systematically — from architecture to integration with infrastructure
• Write in Python, be familiar with PyTorch
• Be ready for cross-functional work with backend and product teams