NeotericLabNEOTERIC LAB / INTELLIGENCE FOR THE NEXT GENERATION
Intelligence for the Next Generation
Expanding the frontiers of intelligence and participation.
From AI learning for the next generation to tailored models,
we bring intelligence into the real world.
01 / OUR PURPOSE
Technology matters for what it makes possible.We believe the next generation should have the opportunity to understand AI, use it, and help create it. Starting with learning opportunities for the next generation, Neoteric Lab works to open AI exploration to more people.
From knowledge to practice, and models to applications, we focus on every step that turns an idea into something real.
02 / RESEARCH & PRACTICE
Across infrastructure, training, evaluation, and deployment,
research and applications advance together.
RESEARCH TOPICS
AI foundations, hands-on learning, and project exploration help juniors turn curiosity into the ability to create.
FRONTIER RESEARCH
Our research into recursive self-improvement (RSI) explores how agents can propose experiments, verify results, and improve models, tools, and learning processes.
Measurable progress informs the next experiment.Plan GPU capacity, clusters, and orchestration for training and inference, connecting compute provisioning with data pipelines and experiment management.
From pre-training and continued pre-training to post-training and alignment. Connect data development, model adaptation, inference optimization, and deployment with domain applications and agent workflows.
Build interactive environments, reward mechanisms, and verifiers for reasoning, tool use, and multi-step tasks. Improve models through cycles of execution, feedback, and training.
Design domain benchmarks, task sets, and judging protocols. Use blind testing, pairwise comparisons, and model arenas to assess capability, cost, and latency, informing model selection and iteration.
03 / FROM IDEAS TO IMPACT
For juniors, developers exploring models,
and teams finding new uses for AI,
we welcome collaborators ready to explore what comes next.
Start with real needs. Define tasks, data, and capability boundaries, then establish measurable goals.
Develop training, environments, and evaluation together, improving capabilities against clear goals.
Integrate models into applications and agent workflows. Use operational feedback and regression evaluation to guide improvements.
OUR MODELS
Exploring efficient intelligence through sparse computation.
From continued pretraining to instruction tuning, Modujo explores mixture-of-experts language models. Public model weights invite further research and experimentation.
Explore Modujo models4B family · Mixture of experts
Continued pretraining / Experimental SFT / 1B base
LET’S BUILD WHAT’S NEXT
Custom models, collaborative evaluation, AI education, or a new research question.
Talk to Neoteric Lab.