G5 Labs
Natural language abstraction for AI-native software development.
The founders
ZW
DP
TK
The company
G5 Labs is building a self-learning, bi-directional compiler that turns natural-language descriptions of software intent into executable code — and works in reverse, mapping existing code into a formal graph of intent, or ontology, that becomes the new source of truth. The platform sits as an application and governance layer on top of AI coding models, letting enterprises resolve merge conflicts semantically, trace code back to its original intent, and enforce policies like GDPR across new and legacy systems.
As AI coding models flood teams with generated code, the harder problem is governing, tracing and modernising it — especially in regulated sectors. G5's bet is that the missing layer is intent, not more generation.
The company is a spinout from MIT CSAIL, built on roughly two years of research led by co-founder and CEO Tim Kraska before its public launch on 15 September 2026.
Kraska is an MIT CSAIL professor who previously co-founded Einblick Analytics (acquired by Databricks) and held roles at Google Brain and Amazon, and he's joined by co-founders Ziniu Wu and Davide Pagano. Investors including Pillar VC and Battery Ventures backed the thesis with $14M in seed funding.
Who else is building in this space
Developing AI world models that simulate physical environments for robotics, testing and other real-world applications
Funded · $500K
Scientific instrument signal data for AI training.
Funded · $125K
Frontier training data for AI and robotics.
Funded · $125K
Voice agents that talk to every employee to map how work gets done and support transformation
Thousands more founded every month.
Dealroom tracks new companies as they appear — founders, funding and signals, over the API or in the platform.