N
Native Tech Stack
Data reasoning infrastructure for structured AI decision-making
Native is a 2025-founded startup building foundation models for tabular data and structured reasoning. The stack—PyTorch, Ray, JAX, Python, Rust, Java—signals a research-forward org focused on training and inference optimization. Active projects span self-supervised learning objectives for tabular data, LLM agent grounding in structured representations, and long-horizon agent behavior simulation. The hiring mix (7 of 8 recent roles at senior level, split between engineering and research) and repeated focus on GPU cluster optimization suggest the team is racing to solve scaling bottlenecks rather than filling out a mature product org.
What Native Is Building
◆Challenges
- Scaling foundation models across huge datasets
- Optimising training runs on gpu clusters
- Optimizing large-scale training runs on gpu clusters
▲Active Projects
- Reasoning over complex tabular data
- Design and build apis used by internal systems and customers
- Scaling foundation models across huge datasets
- Developing new encoders for heterogeneous data
- Developing self-supervised learning objectives for tabular data
- Building scalable training and evaluation pipelines
- Grounding llm agents in structured data representations
- Build training, simulation, and evaluation environments for long-horizon agent behavior
- Productionise research into robust backend services
- Core user interfaces for research tools, internal systems, and customer-facing products
Hiring Activity
8 roles · 8 in 30d
Department
Seniority
Company intelligence
