Rhinovate Raises Pre-A Funding to Accelerate High-Performance Materials R&D With Self-Evolving Physical AI | Unity Ventures Portfolio
Building Reusable AI Infrastructure for Materials R&D
AI-powered new materials R&D company Rhinovate™ recently completed a multi-hundred-million-yuan Pre-A funding round with participation from Unity Ventures.
Rhinovate is a Physical AI-driven new materials R&D company dedicated to transforming complex, experience-dependent materials development into a computable, verifiable, and scalable engineering system.
Unity Ventures believes Rhinovate combines top-tier academic foundations from Peking University and its Shenzhen Graduate School with exceptionally solid industrial execution experience. As an industry pioneer in establishing a full-process wet-dry closed loop, the company has repeatedly earned high recognition from clients and local governments through its hardcore technology. Unity Ventures looks forward to Rhinovate continuing to demonstrate its leadership in AI4M and injecting new momentum into the deep transformation of the materials industry.
AI-powered new materials R&D company Rhinovate™ recently completed a multi-hundred-million-yuan Pre-A funding round, led by CDH Baifu with follow-on investments from Unity Ventures, Binfu Capital, Loongson Venture Capital, Shunxi Fund, and ZGC Capital.
This marks the company's third funding round within a year, following previous investments from Kinzon Capital, YuanSheng Ventures, Shanghai Future Industry Fund, and XtalPi. The proceeds will be primarily allocated to the development of RhinoMat, its foundational large model for materials science, underlying technology R&D, multi-scenario orchestrable AI agents and autonomous evolution laboratory systems, and scale-up validation of carbon-based high-performance materials — continuously deepening core technologies and experimental data assets.
Founded in 2025 and incubated with Peking University's research capabilities, Rhinovate's core team members hail from Peking University, Tsinghua University, University of Cambridge, Stanford University, and MIT. The company is dedicated to transforming complex, experience-dependent materials development into a computable, verifiable, and scalable engineering system.
At its center is the RhinoAI Intelligent R&D Platform, which integrates AI Scientist, autonomous experimentation systems, and pilot-scale amplification capabilities to build an end-to-end R&D closed loop covering "design — execution — characterization — feedback — optimization." This allows models, data, formulations, and processes to continuously evolve through real experiments, pushing high-performance materials R&D from "empirical trial-and-error" toward "intelligent design."
RhinoAI Intelligent R&D Platform: From "Wet-Dry Closed Loop" to "Autonomous Evolution"
New materials are the physical foundation of all hard tech, yet their R&D paradigm has long remained stuck in empirical trial-and-error: from design to mass production typically takes five to ten years, with computation, experimentation, and scale-up operating in silos. Models struggle to translate into processes, data fails to flow back, and performance often degrades during scale-up.
Rhinovate uses cross-scale matter compilation as its underlying unified theoretical framework, making materials R&D a continuously learning, verifiable autonomous evolution platform. Its self-developed Rhino system, with AI Scientist and Self-Driving Lab at its core, builds a Physical AI platform that bridges scientific cognition, experimental execution, and feedback-driven evolution. The platform doesn't merely iterate models autonomously, design material formulations, and recommend equipment parameters — it also autonomously plans experimental protocols, schedules execution workflows, and ultimately completes material preparation and validation in the real physical world.
Rhinovate RhinoAI Intelligent R&D Platform
Rhino Scientist serves as a full-chain research intelligent agent capable of understanding research objectives, calling upon knowledge, materials models, and algorithms to complete knowledge exploration, protocol design, and task orchestration. Multiple agents can be flexibly orchestrated for different materials scenarios, translating scientific problems into executable computational, experimental, and engineering tasks. Rhino Lab, composed of robots, intelligent experimental workstations, high-throughput characterization equipment, and unified control systems, serves as the experimental vehicle through which AI Scientist enters the physical world. It receives research strategies and autonomously completes experimental orchestration, sample preparation, multi-modal characterization, and data feedback within a wet-dry closed loop of computational simulation and experimental calibration, adjusting subsequent experiments based on feedback.
In this system, AI Scientist proposes hypotheses and strategies, robots execute at high throughput, and human scientists define objectives, interpret mechanisms, and make engineering decisions. Together they form a self-driving closed loop of "design — execution — characterization — feedback — optimization," where every experiment drives updates to the knowledge base and models, making the system more precise with each iteration.
Three-Layer Iteration Flywheel: Co-Evolution of Models, Formulations, and Equipment Parameters
On the model side, Rhino is building RhinoMat, a model oriented toward polymers. Built on a foundation of three-dimensional equivariant graph neural networks and long-range interaction modules, it provides unified representation of polymer structures. It constructs force field models to simulate polymer systems, combines diffusion models for inverse generation of target polymers, and incorporates structural representations into surrogate models for Bayesian optimization. Furthermore, the model aligns with experimental data and continuously iterates based on experimental feedback.
On the data side, Rhino has accumulated over 1.25 million multi-scale simulation data points and 48,000 automated experimental data points in structural bearing polymers, olefinic carbon materials, and electrolyte directions, and has built a knowledge graph containing 214,000 entities and 626,000 entity relationships. Regarding experimental data, unlike public databases, Rhino continuously collects full-process data spanning "process — structure — characterization — performance," and retains records of failed experiments. Through data quality control and causal evidence chain construction, it ensures full traceability of data and attributable results.
Each round of real experimental data flows back to calibrate the model; the model then generates and screens candidate formulations for the next round; experimental results in turn drive adjustments to equipment and process parameters, forming a three-level progressive reinforcement loop of "model — formulation — equipment." Continuous mutual feedback between model-side inference and robot-side measurement strengthens model capabilities with each R&D cycle while simultaneously accumulating proprietary data, formulation insights, and process experience.
From "Materials Demonstration" to "Industrial Scale-Up": Building Reusable AI R&D Infrastructure
Spanning the full chain from research to industry, Rhinovate engages in deep collaboration with Peking University scientist teams. Leveraging its presence in Beijing, Shenzhen, and Hong Kong, it converges research capabilities and innovation resources to form a cross-regional collaborative R&D network, conducting joint R&D with industry leaders including Wanhua Chemical, BTR, and JinYu New Energy.
Looking ahead, the company will continue advancing the RhinoMat materials science foundational large model and underlying technology system, with polymer models as the initial direction, building multi-scenario, orchestrable AI scientist systems with autonomous evolution capabilities that bridge computation, experimentation, and industrial application in a closed R&D loop. As more materials directions come online, models, data, experimental skills, and process knowledge will continuously accumulate and become reusable, pushing materials R&D from one-off projects toward platform-based infrastructure — making materials discovery and industrial transformation a computable, verifiable, and scalable sustained capability.
Haifeng Lyu, President of Rhinovate, stated: "Materials are a crucial physical foundation for hard tech. Over the past year, we've brought AI Scientist into real laboratories, used autonomous experimentation systems to make and understand materials, and pushed results toward scale-up validation. In the next phase, this system will prove out across more materials directions, making every experiment the starting point for the next discovery."
