Research Associate in Adaptive and Efficient LLM Architectures

Imperial College London; Greater London; On-site; GBP 51,000 - 59,000; Full time

Benefits offered by this job

Conference travel funding

Compute resources (GPUs)

World-leading institution

Generous pension

Job summary

Imperial College London, South Kensington campus, invites post-doctoral researchers to join the ERC AToM project on Adaptive Tokenization and Memory in Foundation Models. The role focuses on designing, implementing, and publicly releasing efficient LLM architectures with adaptive memory, latent tokenization, and sparse attention.

You will conduct original research in efficient adaptive LLMs, explore long-context reasoning for code, maths and agentic workflows, publish in top venues, and

Qualifications

• PhD in computer science or equivalent.
• Strong track record in top AI/ML/NLP conferences/journals.

Responsibilities

• Designing and developing novel AI architectures.
• Performing retrofitting, post-training, and evaluation of SOTA open-weight models.
• Publishing results in top-tier conferences and journals.
• Implementing research ideas using PyTorch/JAX, HuggingFace transformers/diffusers, and efficient kernels (Triton/CUDA).
• Contributing to research on adaptive memory, latent tokenization, sparse attention, and long-context understanding.
• Contributing to grant proposals and collaborative research initiatives.

Skills

Research excellence

Programming

LLM training

PEFT

CUDA kernels

Pytorch/JAX

PhD holder

Education

PhD in Computer Science

Tools

PyTorch

JAX

HuggingFace

Diffusers

CUDA

Triton

Job description

South Kensington

We’re seeking talented post-docs who have conducted cutting-edge research and/or have extensive experience in frontier AI labs. The position is fully funded by Dr. Edoardo Ponti’s ERC project AToM (Adaptive Tokenization and Memory in Foundation Models) with a focus on designing, implementing, and publicly releasing LLMs with efficient architectures (including adaptive memory, latent tokenization, sparse attention, multi-token prediction, adaptive depth, among others).

This project will lead not only to substantial gains in efficiency (several orders of magnitude speedups without accuracy degradation) but also to the emergence of new capabilities: adaptive FMs can operate over broader effective horizons.

You will conduct original research in the new and exciting field of efficient and adaptive LLM architectures and explore its applications across long-context understanding and reasoning (for code, maths, and agentic workflows) as well as long-horizon multimodal world modelling. We will strive to release new, more efficient and capable AI models and to publish in top-tier conferences and journals. You will work in:

• Designing and developing novel architectures for AI models
• Performing retrofitting, post-training, and evaluation of SOTA open-weight models
• Publishing results in top-tier conferences and journals
• Implementing research ideas using modern deep learning frameworks (PyTorch/JAX), model/dataset libraries (Huggingface transformers/diffusers), and efficient kernels (Triton/CUDA)
• Contributing to research projects on adaptive memory, latent tokenization, sparse attention, long-context understanding and reasoning, agentic, and multimodal world modelling
• Contributing to the life and development of Edoardo Ponti’s Lab, including meetings, presentations, maintaining the website and other resources.
• Contributing to grant proposals and collaborative research initiatives.

What we are looking for:

• You’re expected to have a strong track record in top conferences and journals in the fields of AI/ML/NLP, such as NeurIPS, ICML, ICLR, *ACL, EMNLP, etc. Candidates with research experience as part of frontier AI labs are also welcome.
• Excellent skills in coding, strong foundations in mathematics, and knowledge in deep learning.
• Practical experience in a broad range of techniques including LLM training, evaluation, RLVR, PEFT, quantisation, tensor/data parallelism.
• Ideally, familiarity with CUDA kernels and/or Triton, and inference engines (vLLM, SGLang, et cetera).
• Experience coding with deep learning libraries such as Pytorch/JAX is essential.
• Applicants must hold a PhD in computer science or equivalent.

See job description for full requirements.

What we can offer you:

• Extensive funding for conference travel (2 international conferences per year)
• Access to extensive compute via the AToM project GPUs (B200s and cloud credits) and GPUs (A100s and H200s)
• The opportunity to continue your career at a world-leading institution
• Sector-leading salary and remuneration package (including 43 days off a year and generous pension schemes).

Full-time, fixed term post for 2-years to start winter 26/27 (flexible)

*Candidates not yet awarded their PhDwill be appointed as Research Assistant within the salaryrange£45,399- £48,876per annum.

Closing Date: 30 September 2026 (midnight BST)

£50,733 to £59,484 per annum