Data and go-to-market infrastructure for robot foundation models
Poke & Wiggle builds tooling to address data collection and commercialization bottlenecks in robot foundation models. The stack—JavaScript, Python, C++, Rust, gRPC, Kubernetes, and WebXR/OpenXR—signals a systems-level approach to real-world robot data capture and simulation, with heavy infrastructure (Bazel, Nix, ArgoCD, Flux) suggesting they're solving reproducibility and scale challenges. Active projects around multi-robot fleet scaling, dexterous manipulation (two-handed tasks), and physical data workflows indicate they're moving beyond single-robot setups into operations-grade problems.
Poke & Wiggle, founded in 2025 and based in Munich, is a 2–10 person team tackling infrastructure for robot foundation models. Their focus spans two areas: the data layer (collecting, labeling, and simulating robot interactions at scale) and the go-to-market layer (how models trained on that data get deployed and monetized). The engineering-heavy composition (4 engineers, 2 ops roles) with a mix of lead and junior levels suggests both hands-on execution and knowledge transfer. Active hiring signals growth across the stack. Current operational challenges center on fleet throughput, consumables management, and scaling across multiple robots—typical pain points in physical-world AI systems where data collection is the bottleneck.
Core languages: JavaScript, TypeScript, Python, Rust, C++. Infrastructure: Kubernetes, ArgoCD, Flux, Bazel, Nix. Observability: Datadog, Honeycomb. Simulation and XR: WebXR, OpenXR, Three.js, A-Frame, Meta Quest. Robotics: ROS 2, WebRTC, gRPC.
Multi-robot fleet scaling, dexterous manipulation (two-handed robot tasks), physical data collection workflows, procurement and inventory management, and workflow architecture. Core challenges are high-throughput performance, fleet coordination, and managing consumables.
Other companies in the same industry, closest in size