MindWalk Brings OpenFold3 into Production in Minutes with AMD Inference Microservices on Vultr Kubernetes Engine

Antibody-antigen inference ran approximately 5x faster than in MindWalk's prior environment. The deployment, run with engineering partners AMD and Vultr on AMD Instinct™ MI325X GPUs, demonstrates the model-agnostic architecture and elastic compute beneath ReefIQ™ as MindWalk onboards enterprise partners.

MindWalk Holdings Corp. ("MindWalk") (NASDAQ: HYFT), a Bio-Native AI company, today announced results from a production deployment of the open-source OpenFold3 structure prediction model, run with engineering partners AMD and Vultr on Vultr Kubernetes Engine and Vultr Cloud GPU powered by AMD Instinct™ MI325X GPUs and delivered through AMD Inference Microservices (AIM). It also served as an independent validation of AIM for OpenFold3 in a production life-sciences setting. MindWalk designed the evaluation around its own production workloads, spanning protein, RNA, DNA, ligand, and antibody-antigen prediction, and ran the complete path from preprocessing through inference inside its discovery workflows, under its governance and human scientific review. A production-grade prediction environment deployed in minutes, and antibody-antigen inference time fell by approximately 5x against MindWalk's prior environment. The deployment expands the compute capacity available beneath ReefIQ™, MindWalk's biological context layer, as the company scales enterprise adoption following ReefIQ's June 2026 launch. Full methodology and results are published in the case study at https://discover.vultr.com/mindwalk-vke-use-case.

"We chose an open-source model deliberately," said Dr. Jennifer Bath, President and Chief Executive Officer of MindWalk. "Predicting the shape of a protein is becoming something any team can buy. Knowing what that shape means for a disease, and for the programs a company has already run, is not. That is what ReefIQ holds. What this deployment changes is how quickly and how cheaply we can bring a new enterprise partner onto it. AMD and Vultr have turned the computing side into a standard path, so a partner's programs can be running in minutes instead of waiting on a custom build. That shortens the time from signing to first result, and it means our costs do not have to climb with every partner we add. We tested it on real antibody-antigen programs rather than a set chosen to look good, because that is the test our partners apply."

What MindWalk ran

MindWalk ran single- and multi-chain proteins, protein-RNA and protein-DNA interactions, protein-ligand structures using CCD and SMILES inputs, and antibody-antigen complexes drawn from its own discovery programs, exercising the complete path including CPU-intensive preprocessing and multiple sequence alignment alongside GPU-accelerated inference.

The evaluation also surfaced practical considerations around multi-GPU execution, inference controls, and deployment workflows. MindWalk shared those findings with AMD and Vultr to inform subsequent platform work.

“MindWalk is showing what’s possible when open-source AI is applied to complex biomolecular structure prediction to advance drug discovery. Together with AMD, Vultr provides the Kubernetes-native CPU and GPU infrastructure behind that work, helping teams spend less time managing infrastructure and more time advancing scientific discovery with models like OpenFold3.” — Kevin Cochrane, Chief Marketing Officer, Vultr

Why this matters for the industry

Prediction capability is standardizing quickly. Model weights are open, silicon vendors package the models as deployable microservices, and a marketplace transaction now delivers in minutes what took a well-funded research organization years to assemble. For pharmaceutical R&D organizations operating under an AI mandate, that removes one of the more expensive obstacles between a research question and an answer.

What standardization does not supply is the biological context a model reasons over, or the governance that makes its output usable inside a regulated program. That is the layer ReefIQ™ provides and it grows more valuable with each program run on it.

What the validation establishes

The model-agnostic architecture is demonstrated, not asserted. OpenFold3 is not a MindWalk model. MindWalk deployed, ran, and measured a third-party open-source model on AMD Instinct GPUs in the Vultr cloud, inside its own discovery workflows and under its own governance and human scientific review. Partners that bring their own models onto ReefIQ™ follow the same path.

Compute efficiency is a cost lever. Vultr and AMD form the compute fabric beneath MindWalk's platform. Inference efficiency at this layer bears directly on the cost of delivering AI workflows to partners at enterprise scale.

Deployment friction is falling for partners, not only for MindWalk. The standardized, Kubernetes-native path from open model to production workflow that AMD and Vultr have built shortens the distance between a research question and an answer inside a regulated environment. That is the operational objection pharmaceutical CIOs raise most often, and it is the objection that sets the pace of enterprise onboarding onto ReefIQ™.

Results reflect the specific configurations evaluated and are not representative of all workloads.

Availability

ReefIQ™ and LensAI™ are commercially available from MindWalk. AMD Inference Microservices for OpenFold3 are available through the Vultr Kubernetes Engine Marketplace as part of the AMD Enterprise AI Software stack.

About MindWalk Holdings Corp.

MindWalk Holdings Corp. (NASDAQ: HYFT) is a Bio-Native AI company building the BioIntelligence infrastructure that life sciences AI and agentic AI require, integrating AI, data, and advanced wet lab capabilities into one connected discovery ecosystem. At its core is HYFT^® Technology, a proprietary, function-aware representation of biology. Its HYFT fingerprints span sequence and structural biology and, refined over 20 years of curation, form a continuously evolving biological representation of 660 million patterns and 25 billion relationships. This enriched biological representation is the architecture behind ReefIQ™, the biological data substrate that provides context for life sciences, enriching data at ingestion and growing more valuable with each program run on it, and LensAI™, the reasoning and application layer for target discovery, candidate diligence, portfolio decision support, and the agentic AI workflows pharma is racing to deploy. By design, value compounds in this HYFT representation layer, not in any individual AI model that runs on top of it.

Forward-Looking Statements

This press release contains forward-looking statements within the meaning of applicable United States and Canadian securities laws. Forward-looking statements are based on management's current expectations and assumptions and are subject to known and unknown risks, uncertainties, and other factors that may cause actual results to differ materially, including: the pace and scale of enterprise adoption of ReefIQ™ and LensAI™; statements regarding the continued standardization of AI model capability and deployment, which may not develop as anticipated; the technical performance of AI-based structure prediction methods and of the underlying compute infrastructure; the fact that the inference results described reflect specific evaluated configurations and may not be representative of other workloads, inputs, or environments; reliance on third-party cloud, GPU, microservice, and open-source model providers, and the continued availability and pricing of those resources; the multi-GPU execution, inference control, and deployment workflow considerations identified during the evaluation and the timing of any resolution; the ability to convert technical validation and engagement into contracted, recurring arrangements; the build-out and certification of compliance capabilities for regulated workloads; the ability to enter into pharmaceutical, infrastructure, or other arrangements; competition; intellectual property risks; regulatory determinations; and capital markets conditions. Additional information is available in MindWalk's Annual Report on Form 20-F and other filings on SEDAR+ (sedarplus.ca) and EDGAR (sec.gov/edgar). Except as required by law, MindWalk undertakes no obligation to update any forward-looking statement. Nothing in this press release constitutes investment advice. This press release does not constitute an offer to sell or the solicitation of an offer to buy any securities.

Trademarks

HYFT^® is a registered trademark of MindWalk Holdings Corp. LensAI™ and ReefIQ™ are trademarks of MindWalk Holdings Corp. or its subsidiaries; ReefIQ™ registration is pending. All other trademarks are the property of their respective owners.

Source: MindWalk Holdings Corp.

View source version on businesswire.com: https://www.businesswire.com/news/home/20260921650812/en/

AMD and Vultr have turned the computing side into a standard path, so a partner's programs can be running in minutes instead of waiting on a custom build