July 29 (Wed), 14:00~: From "Using As-Is" to "Using Together": Webinar on MACE-Osaka Fine-tuning Practices

QunaSys Inc. (Headquarters: Bunkyo-ku, Tokyo; CEO: Tenko Yamasaki), which promotes the industrial application of quantum computers, will hold a webinar online on July 29, 2026 (Wed) from 14:00 to 15:00, with the theme "From 'Using As-Is' to 'Using Together': Practices of MACE-Osaka Fine-tuning".

In recent years, machine learning-based potentials (MLIP: Machine Learning Interatomic Potential) have rapidly developed in the fields of materials development and molecular simulations, enabling significant reductions in computation time compared to conventional first-principles calculations (DFT). On the other hand, many users report challenges such as insufficient accuracy for their specific research themes when simply using general-purpose models as-is, or uncertainty about how to evaluate them.

This webinar will explain the benefits MLIP can bring to R&D, using MACE-Osaka as an example. It will also specifically demonstrate how fine-tuning (a method of adjusting an existing model with proprietary data) can improve accuracy, using the example of adsorption energy on a copper surface.

Webinar Overview

Event Name

From "Using As-Is" to "Using Together": Practices of MACE-Osaka Fine-tuning

Date and Time

July 29, 2026 (Wed) 14:00-15:00

Format

Online (Zoom)

Participation Fee

Free

Speakers

Wataru Minakami, Professor (Osaka University, Center for Quantum Information and Quantum Biology, Ph.D. in Science)

Shunya Onishi (QunaSys Inc., Research Solution Department)

Background and Objectives of this Webinar

The advent of general-purpose MLIPs is creating an environment where anyone can use high-performance models. However, decisions such as which model to choose, how much to adjust it with proprietary data, and what constitutes "usable" are increasingly left to the user.

In this webinar, Professor Minakami will first provide an overview of MLIPs and their current position, and introduce "MACE-Osaka," which is currently attracting attention. He will also discuss how these technological trends will be related to quantum computers in the future.

Furthermore, based on these medium- to long-term prospects, we will discuss specific use cases of fine-tuning from the perspective of MLIP adjustment (adjustment) that can be implemented immediately.

Program (Tentative)

14:00–14:05 Introduction

14:05–14:35 Introduction to MACE-Osaka and Future Connections with Quantum Computing (Prof. Minakami)

14:35–14:45 Q&A

14:45–14:55 Fine-tuning Case Study (Onishi)

14:55–15:00 Q&A and Closing

*Speakers and content are subject to change without notice.

Highlights of this Webinar

・Features of MACE-Osaka and its practical value for R&D

Discussing the latest trends in MLIP, we will explain the "practical benefits" that the use of MACE-Osaka brings to the actual R&D process in terms of calculation speed and accuracy.

Scenario for collaboration with quantum and high-precision computing for the future We will introduce an approach that complements "weak areas" with quantum and high-precision computing, rather than replacing all general-purpose models. We will discuss prospects for future fine-tuning and calibration data design.

・Practical Demo: Fine-tuning adsorption energy on a copper surface

A calculation demo will be performed using the adsorption of molecules on a copper surface as a subject. The fine-tuning results of MACE-Osaka will be compared with DFT (first-principles calculations) and other general-purpose models (UMA, etc.) to organize the differences in accuracy and calculation time.

・Evaluation design directly linked to practical work, not just "accuracy somehow improved"

We will explain practical evaluation methods that go beyond simple energy curve improvements and are directly linked to decision-making for "next actions" in actual materials development, such as structural relaxation, forces, and MD stability.

Recommended for:

Those who want to try MLIP but are unsure what to evaluate for their own themes

Those who prioritize accuracy usable for research decisions, not just calculation speed

Those who want to know if their proprietary data can be used for fine-tuning

Those who are struggling with the balance between general-purpose models and in-house optimization

Those considering a computational infrastructure that includes quantum computing in the future

How to Apply

Please register in advance via this form.

*Application deadline: July 27, 2026 (Mon)

*The viewing URL will be sent by email the day before the event.

Post-Webinar Information

After attending this webinar, the following consultations are available depending on your interest:

PoC (Proof of Concept) for fine-tuning using MACE-Osaka

Individual consultations on evaluation design for MLIP/NNP

Consideration of medium- to long-term roadmaps, including quantum and high-precision computing

Company Profile

QunaSys is a quantum computer software company leading the development of quantum algorithms for industries such as chemistry and materials science. Focusing on quantum machine learning, quantum chemistry, and CAE (Computer-Aided Engineering) using quantum technology, we collaborate with academic institutions, industry, and government agencies to maximize the potential of quantum computers and tackle scientific challenges.

Contact Information

QunaSys Inc. Webinar Desk E-mail: [email protected]