AI development
Foundation-model tooling, applied AI products, and the developer infrastructure that turns research into deployed systems. We look for teams shipping into real workflows — not demos.
From research to production
Model & developer tooling
Fine-tuning, orchestration, and observability — the toolchain that makes working with models feel like software engineering instead of alchemy.
Applied AI products
Vertical products that embed intelligence into real workflows — legal, healthcare, industry — where the moat is workflow depth, not the model.
Inference infrastructure
Serving, routing, and cost optimization. As inference becomes the dominant compute bill, efficiency becomes a product category of its own.
Data engines
Curation, labeling, and synthetic generation pipelines. Quality data is the scarcest input in AI — and the least glamorous to build.
Shipping beats demoing
The gap between an impressive demo and a dependable product is where most AI companies die. We back the ones engineering their way across it.
Production deployments with measurable workflow outcomes
Unit economics that improve as models get cheaper
Evaluation discipline — teams that measure before they claim
Model-agnostic architecture — no single-vendor dependency