The AI Research Residency is a paid 3 to 6 month program for late-stage PhD students and exceptional early-career researchers who want to pursue high-impact AI research grounded in real systems.
Residents work closely with Maincode’s research and engineering teams, with dedicated access to large-scale GPU compute and our in-house research infrastructure. You will explore open problems, run experiments at scale, and produce work that can contribute to top-tier publications, open research, infrastructure, or the systems behind Matilda.
This role is research-first, but applied. Strong projects may take the form of model research, evaluations, infrastructure, technical engineering work, or product-facing research that improves Matilda and future Maincode systems.
We are interested in research that makes AI systems more capable, reliable, efficient, and useful in the real world.
The residency program is primarily focused on the following areas:
•
Tool use, planning, memory, computer control, multi-agent systems, and safe execution in real-world environments.
•
Long-context reasoning, workflow understanding, state tracking, memory systems, and methods for maintaining coherence across complex tasks.
•
Capability evaluations, alignment, oversight, interpretability, robustness, red-teaming, and benchmarks for real-world task completion.
•
Language model training, reinforcement learning, reasoning methods, optimisation, architectures, and new approaches to improving model behaviour.
•
Data curation, filtering, synthetic data, mixture design, quality verification, pruning, and principled approaches to training signal.
•
Vision-language models, grounding, perception, multimodal reasoning, and systems that combine language, visual context, and action.
Efficiency and infrastructure
•
Training and inference efficiency, kernels, parallelism, sharding, decoding, quantisation, scheduling, and systems for running large models reliably.