Research Intern - Multi-Agent Systems
Microsoft
Multi-agent AI systems are driving the growth of intelligence scaling and real-world impact. At Microsoft Research Cambridge, future AI infrastructure (FAI) team is looking for two passionate research interns to explore future multi-agent AI systems together. This research internship is about ML and systems co-design. To build better multi-agent AI, we will explore what new capabilities are needed for future agent models, such as recursive LLMs, latent memory, and task planning. At the same time, we will look into innovative system designs like better agent-level parallelism, agent communication intensity, and distributed memory systems. With FAI team’s AI hardware innovations (see AOC and MOSAIC), the outcome of this work will lead to strong impact on future AI systems design through the collaboration with other MSR research teams and Microsoft product teams.
Responsibilities
Work on designing, training, fine-tuning modern LLM architectures using Pytorch and other relevant tools on GPU clusters.
Work on post-training open-source models into agentic and/or recursive variants with emerging capabilities such as planning, state-tracking, and improved latent memory, potentially by using gradient-free methods such as evolution strategies
Work on multi-agent design and implementation, integrating system-level optimizations (e.g. shared latent KV-cache, agent-level parallelism) using relevant framework tools and inference backends like vLLM and SGLang.
Qualifications
Required/Minimum Qualifications:
Being enrolled in a PhD program of computer science, artificial intelligence, machine learning, computer engineering, electrical and electronics engineering, or other related fields.
Other Requirements:
Experience working on research projects related to ML and systems design.
Experience in model post training, reinforcement learning / evolution strategies, or supervised fine tuning.
Experience in building high-performance LLM inference systems using SGLang or vLLM.
Preferred/Additional Qualifications:
Publications in top ML conferences and/or systems conferences
This position will be open for a minimum of 5 days, with applications accepted on an ongoing basis until the position is filled.
Microsoft is an equal opportunity employer. All qualified applicants will receive consideration for employment without regard to age, ancestry, citizenship, color, family or medical care leave, gender identity or expression, genetic information, immigration status, marital status, medical condition, national origin, physical or mental disability, political affiliation, protected veteran or military status, race, ethnicity, religion, sex (including pregnancy), sexual orientation, or any other characteristic protected by applicable local laws, regulations and ordinances. If you need assistance with religious accommodations and/or a reasonable accommodation due to a disability during the application process, read more about requesting accommodations.