Decart
Data & simulation
Tel Aviv frontier lab building real-time video world models (Lucy, Oasis) plus a cross-silicon optimisation stack.
Decart is a frontier AI lab building real-time generative world models and the optimisation stack that makes them cheap enough to run live. Lucy is a real-time video model for world editing, deployed across virtual try-on, live advertising, social platforms and gaming; Oasis is an interactive world-generation model aimed at physical AI — physically responsive environments for robotics, autonomous vehicles, manufacturing and drones. Underneath sits the Decart Optimization Stack, which runs these models across NVIDIA GPUs, AWS Trainium and Google TPU. The company claims its models are roughly 100x more compute-efficient than conventional approaches, the basis for its thesis of persistent intelligence without persistent compute. Founded in 2023 in Tel Aviv.
Research mapping
Physical AI / Sim-to-Real, Digital Twins & Validation (L9) / Data & simulation
Evidence boundary
Original analysis and independent corroboration remain under development. Commercial scale and durable technical advantage remain under review.
What makes it defensible
A cross-vendor optimisation stack that runs its own world models on NVIDIA, Trainium and TPU at real-time latency — the efficiency engineering, not the model weights, is the moat.
Position in the stack: Simulation and world-model layer: Oasis supplies interactive, physically responsive environments for robot and AV training, competing with conventional simulators on photorealism and on-demand generation.
Who buys it
Unknown
Named customers: Amazon (AWS Trainium deployment)
Founding team
Dean Leitersdorf — Co-founder & CEO — Technologist and PhD; served in Israel's Unit 8200; the product and research lead of the company.
Moshe Shalev — Co-founder — Met Leitersdorf while serving in Unit 8200; the business-side co-founder, keeping the research agenda commercially grounded.
SourceFinancing detail
- Last round
- Series B
Investors: Radical Ventures, NVIDIA, Sequoia Capital, Benchmark, Adobe Ventures, Toyota Ventures, Atreides Management, Valor Equity Partners, Zeev Ventures
Mapped funds and investors
Direct relationships supported by published evidence.
Reported holdings as-of
No sourced relationships mapped yet.
Disclosed investment history
No sourced relationships mapped yet.
Showing up to four relationships in each group. ETF weights describe the fund’s portfolio, not ownership of this company. Private investment history does not establish current ownership.