Overview Miraxis is building the rights-cleared data factory for robotics and physical AI. Post-training at Miraxis is where dataset improvements become measurable model improvements: targeted data expansion, labeling strategy, adaptation experiments, and evaluation. As Chief Researcher (Robotics / Physical AI), you will set the applied research vision across pretraining, post-training/evaluation, and simulation/syntheti data—and ensure the work is operationalized into the product and delivery engine. This role is currently On hold , but we keep it as a leadership anchor for when we’re ready to staff a dedicated research org. Miraxis was founded by Artem Sokolov, who is also the Founder and CEO of Humanoid AI. https://thehumanoid.ai/te%d0%b0m/ What you’ll do (Responsibilities) Define the research vision and technical agenda across: Pretraining data requirements and quality dimensions Post-training loops (adaptation, labeling strategies, evaluation harnesses) Simulation and synthetic expansion strategies with defensible validation Create a measurable framework connecting dataset attributes → evaluation outcomes → customer value (acceptance gates, benchmarks, reporting). Translate research findings into platform requirements (schemas, metadata, validation checks, tooling) in partnership with Product and Engineering/Platform. Guide the build-out of a small, high-leverage research team: hiring plan, mentorship, and collaboration model with engineering/product/ops. Engage directly with customers/partners as the technical research authority, shaping specs and evaluation standards. What we’re looking for Track record leading applied robotics / physical-AI research and shipping outcomes (methods, benchmarks, systems), not just publications. Strong judgment on what matters for modern VLAs/VLMs (data attributes, evaluation, sim-to-real) and ability to translate into roadmaps and standards. Comfort operating in an early-stage environment: ambiguity, rapid iteration, and high cross-functional ownership. Excellent communication and ability to lead a remote/hybrid team. Nice to have Experience building evaluation frameworks used in production (not only academic benchmarks). Familiarity with multimodal robotics data ecosystems (logs, synchronization, metadata, provenance). Prior experience working with enterprise/R when opened, this will be a high-ownership leadership role with deep cross-functional engagement. Remote/hybrid collaboration with occasional travel for key partner/customer work. Location Remote-friendly. See the Location property for current constraints.
