Nebius PhD Fellowship Program

Making large-model serving faster, cheaper, and more scalable on modern GPU clusters.

Proposals welcome on:

• speculative decoding
• KV-cache compression and management
• low-precision inference (FP8/FP4 quantization)
• sparse and long-context attention
• disaggregated prefill/decode architectures
• MoE serving
• kernel- or scheduler-level optimizations

Strong projects pair algorithmic ideas with real systems benchmarks: latency, throughput, and cost per token at scale.

Turning pretrained models into capable, aligned, and reliable systems.

Directions include:

• RL from human or AI feedback
• RL with verifiable rewards
• reward modeling
• preference optimization (DPO and beyond)
• reasoning and chain-of-thought training
• synthetic data generation and distillation
• rigorous evaluation of post-trained models

Work that improves sample efficiency or makes post-training pipelines more reproducible and open is especially valued.

LLM-based agents that plan, use tools, and act over long horizons.

Topics include:

• multi-agent orchestration
• tool-use and code-execution agents
• memory and long-horizon planning
• RL environments and training for agents
• agent benchmarks and evaluation
• reliability and safety of autonomous systems

Proposals should target measurable improvements on realistic tasks (e.g. software engineering, research automation), not demos alone.