#None #San Francisco #AI Training #Mind Lab #Mint Recursive

Mind Lab Launches Mint Recursive for Enterprises

Mind Lab has unveiled its innovative platform, Mint Recursive, designed for businesses looking to develop industry-specific AI models. This new offering combines essential aspects of model training, including benchmark design, data management, post-training processes, and deployment into a single workflow. The launch aims to assist organizations across various sectors—ranging from pharmaceuticals to gaming—in transforming their proprietary knowledge into advanced AI capabilities.

The Core Offerings of Mint Recursive

In any engagement, the process begins with developing a custom benchmark alongside the client. This benchmark helps define the objectives, expectations, and common challenges, ensuring that the AI model aligns precisely with the company’s requirements. By integrating data development with Large Language Model (LLM) post-training, Mint Recursive operates on a continuous feedback loop where each iteration is evaluated against the benchmarks set by the client.

Cost-Effective Solutions

Mind Lab recognizes that expenses often hinder companies from creating their own AI models. To address this, they've optimized the Mint Recursive platform using Low-Rank Adaptation (LoRA) post-training, shared foundational models, and distributed optimization techniques. This ensures that less computing power and fewer parameters are needed, which has reportedly decreased GPU costs on executed projects by more than 90%. With introductory training sessions starting at just $1,000, businesses can access a range of base models—including those from GLM and Qwen families—capable of being deployed in private or hybrid environments.

Flexibility in Implementation

Organizations using Mint Recursive have options tailored to their team sizes and capabilities. Mind Lab’s Forward-Deployed Engineers (FDEs) can manage the entire process from benchmark development to model deployment. Alternatively, teams can utilize the platform's automation features to oversee model definition, data management, and performance assessment. For those demanding full control, Mint Recursive allows developers to customize their training loops and loss functions while benefiting from the lab’s infrastructure, including trajectory sampling and weight synchronization.

The Foundations of Innovation

The Mint Recursive platform builds upon Mind Lab's extensive research in reinforcement learning. The lab was a pioneer in executing end-to-end LoRA reinforcement learning on extensive models with trillions of parameters. Within weeks of launching Mint Recursive, it also introduced the Macaron-V1.1 model, demonstrating an impressive increase in performance, validating the efficiency of its post-training capabilities.

In the healthcare and life sciences fields, Mind Lab has shown the effectiveness of this technology by supporting activities such as clinical literature review, drug discovery, and ensuring quality in medical charting. The company has achieved success across various quality control metrics, showcasing the transformative impact of combining expert judgment with advanced AI.

Chen Kaijie, the founder of Mind Lab, emphasized the importance of leveraging in-house expertise for training AI models. He stated, "An enterprise's most valuable training signal is the judgment of its own experts. That judgment lives in the reviews, corrections, and decisions a general model never sees. Post-training is how it becomes part of a custom AI model the company owns. It doesn't need to start as a big program. It can start with one real task."

Conclusion: A Call to Action for Enterprises

Organizations prepared to embark on their AI development journey can join Mind Lab and begin with identifying a singular task to base their initial model development. Through Mint Recursive, companies are empowered to transform their specific industry knowledge into tailored AI models that enhance operational efficiency and drive innovation.

For more information about Mint Recursive and how it can benefit your organization, visit Mind Lab’s website.