Researching what comes after today's AI.
VeloceLabs is the research arm of VeloceAI, exploring new approaches to intelligence, efficient computing, autonomous systems and emerging technologies. We turn promising research into experiments, open-source technology, prototypes and eventually real-world products.
Our research labs
Five focused programs, each exploring a different frontier. Early stage, coherent thesis, real work underway.
Researching smaller, more capable and more efficient intelligent systems, from specialised language models to reasoning, memory and autonomous agents.
Exploring the systems and computing architectures required to make advanced AI faster, cheaper and more scalable.
Bringing intelligence into the physical world through robotics, embodied AI, sensing and next-generation materials.
Exploring how artificial intelligence can help scientists generate hypotheses, run simulations and accelerate discovery.
Exploring computation beyond conventional architectures, from quantum systems to neuromorphic and photonic approaches.
Current research
Real problems we're working on now, not a list of futuristic topics.
Efficient Small Language Models
Develop specialised models that can perform enterprise tasks while requiring substantially less compute and inference cost.
Long-Context Intelligence
Research approaches for reasoning over extremely large information spaces without simply increasing transformer context indefinitely.
Expert Intelligence
Explore architectures where specialised expert models or modules collaborate dynamically rather than relying on one enormous general-purpose model.
Faster AI Inference
Investigate model optimization, inference techniques and alternative computing approaches for lower latency and lower cost.
Autonomous Agents
Build AI systems capable of reasoning, planning, memory, tool use and completing longer-running tasks with appropriate human control.
Model Evaluation & Reliability
Research methods for measuring behaviour, migration safety, model quality, hallucination, cost and performance across different AI systems.
From research to real technology
VeloceLabs investigates difficult technical problems before they become obvious product categories. Successful research can become open-source infrastructure, published work, intellectual property or technology incorporated into VeloceAI products.
Future frontiers
Areas we intend to grow through internal research, researchers, university collaboration and industry partnerships.
Future directions we're building toward, not laboratories we operate today.
Research philosophy
BUILD, DON'T JUST READ
We implement ideas, benchmark them and learn from real systems.
MEASURE EVERYTHING
Research claims should be testable, reproducible and supported by evidence.
RESEARCH → REALITY
The best research should eventually create technology that matters outside the laboratory.
Build the future with us.
We are looking for unusually curious students, researchers and engineers interested in difficult problems across AI, computing, robotics and emerging technology.
Exceptional applicants may be invited to interview for research internships, engineering roles or project-based research opportunities. We can't guarantee selection.
Apply to VeloceLabsResearch with us.
We are open to research collaborations, compute partnerships, academic partnerships and conversations with organizations working on frontier technology.
Partner with VeloceLabs