Summary

• Protege offers multimodal data aggregation and standardization with AI training/evaluation curation aligned to foundation model builders, delivered with enterprise-grade compliance and repeatable, deal-driven programs.
  
  

• Segmed offers multimodal data access through direct connections within its broad healthcare provider network (all 50 U.S. states & international) focused on imaging data with multimodal extensions, provides real world evidence analytics and insights through its in-house clinical team based on specialized oncology, neurology, and cardiovascular cohorts, and enterprise readiness through highest security, compliance, and privacy standards (ISO 27001, SOC 2, Expert Determination).
  
  

• Protege and Segmed combined unlocks a single, AI developer–ready multimodal package that unifies Segmed’s imaging depth and compliance with Protege’s cross-modal standardization and customized AI builder data delivery needs. This accelerates delivery for AI data buyers while expanding a growing revenue stream for the data partner, avoiding the overhead of building an AI-specific commercialization team and preserving flexibility in their broader AI strategy.
  
  

The Data That Today’s Healthcare AI Companies are Searching For

AI developers increasingly expect integrated, multimodal datasets that reduce time-to-value and clear enterprise hurdles. They’re no longer satisfied with disconnected raw feeds; they want their teams to be able to focus on model development, rather than data wrangling.

Instead of juggling multiple healthcare providers, contracts, and schemas, buyers are looking for the full package, not just raw materials. In practice, that means:

• Single standardized datasets spanning multiple modalities that often include imaging plus complementary data such as EHR and pathology slides

• Consistent metadata and governance across sources

• One contract, one security review, and cohesive delivery rather than fragmented point solutions

All of this has to be enterprise-ready from day one. AI developers need offerings that can support sensitive protected health information (PHI) workflows and adapt to buyer-specific security requirements.

Scale is another non-negotiable. Foundation models demand large, diverse volumes, but not “volume for volume’s sake.” Buyers expect:

• Big, heterogeneous datasets suitable for high-capacity model pretraining

• Hard-to-source cohorts, such as oncology subtypes and rare conditions, to address sparsity where it most affects model quality

Finally, they care deeply about repeatability. They want curated datasets that can be refreshed and extended across programs, with:

• Clear SLAs and transparent data provenance

• Consistent delivery mechanisms across teams and use cases

• Clear licensing terms and data security expectations so they can plan multi-year roadmaps on a stable commercial and governance foundation

• Direct integration with IT systems of healthcare providers allowing for live updates of data

This combination of data, scale, speed, and repeatability provides the basis for Protege and Segmed’s partnership to deliver some of the most advanced AI training-ready datasets available in the market today.

The Segmed x Protege Difference 

Together, Segmed and Protege provide imaging depth plus cross-modal unification, enabling AI developers to move faster from concept to production with enterprise-ready datasets. This helps to unlock use cases such as multi-modal foundation model training, AI model development for detection and triage, and AI for workflow automation.

Segmed brings large-scale, specialized imaging data combined with contextual clinical EHR data drawn from direct connections to IT systems within a broad and diverse network of healthcare providers, with coverage across all 50 U.S. states. Protege then turns that breadth into a cohesive product by:

• Connecting imaging with more specific EHR, pathology, and other modalities across its network of data providers

• Standardizing schemas, labels, and metadata so that downstream teams interact with a single dataset rather than fragmented feeds

This combination is enterprise-ready from ingestion through delivery. Segmed has the highest security, compliance, and privacy standards as shown by long-standing ISO 27001, SOC 2, and HIPAA Safe Harbor and Expert Determination, as well as a track record of passing audits with big pharma and top medical device companies. 

Protege layers on secure workflows and delivery patterns specifically tailored for leading foundation model builders, while combining with other data sources in a way that’s tailored for AI training and evaluation workflows. For AI model builders and developers, that means fewer surprises along the data delivery path from procurement to production.

The Protege Data Lab also works directly with leading AI researchers and model builders to design and curate datasets that are purpose-built for AI. This can range from small cohorts to cover specific use-cases to large, diverse cohorts used more broadly for AI pretraining runs. Regardless of the situation, Protege’s team of data experts and researchers work with model builders to create the best possible datasets to meet the specific training or evaluation needs.

The Segmed and Protege partnership is also built around practical multimodality rather than theory. Imaging-centric datasets can be used for AI pretraining use cases where volume is important. Then in the future for follow-on opportunities, connected EHR and pathology support potential fine-tuning and evaluation applications. These overlapping cohorts are integrated across modalities and healthcare providers, delivering them as one unified, researcher and developer-ready package.

Segmed’s Areas of Differentiation

Segmed brings a set of capabilities that make the combined Segmed and Protege offering especially compelling for AI developers who are looking for imaging data solutions at scale.

On the data side, Segmed offers specialized access that goes far beyond generic imaging aggregation. Its diverse network includes small to large health systems, academic and non-academic centers, oncology, neurology, and cardiovascular-focused centers and other high-signal cohorts that are difficult to access via general-purpose aggregators. 

Coverage spans all 50 U.S. states, with international availability in APAC, Europe, and LATAM via partners. Direct integrations with IT systems within U.S. providers support faster and more predictable fulfillment on a wide range of imaging requests. Additionally, direct connections allow for live updates of de-identified datasets.

Segmed’s in-house clinical team is another major difference-maker. Instead of just providing data, Segmed also provides real world evidence analytics and insights based on the data, providing an additional layer of value to researchers. The clinical team also conducts thorough feasibility studies allowing for complex inclusion and exclusion criteria. Rather than limiting cohort design to PACS-level variables, the team can work with richer logic that incorporates elements such as:

• Medication histories

• Risk factors

• EHR-dependent attributes and clinical context

This level of sophistication helps AI developers design cohorts that more closely mirror real-world patient journeys and satisfies partner organizations that require clinical expertise to validate inclusion and exclusion criteria.

Trust is foundational to everything the Segmed team does, reflected in its long-standing track record of security, compliance, and privacy standards. The de-identification and data standardization technology is scalable and works across text, pixel, and meta-data based information. In addition to ISO 27001 and SOC 2 certification, Segmed successfully secured independent Expert Determination across structured clinical data, imaging pixels, and radiology reports. Annual audits are completed with a proven record of clearing demanding reviews for large pharma and medical device companies. All of this ensures responsible data stewardship and that the data curated is ready for AI training and evaluation use cases, without compliance issues. 

Operationally, Segmed sources through direct IT-connections with their own provider network and combines that with their in-house clinical expertise to help curate data effectively for their specific use cases. They also have built-in connections with organizations such as Advocate Health and Fox Chase Cancer Center to access pathology, radiology, and EHR streams that enrich the imaging foundation. 

Beyond being a data provider alone, the team collaborates with major academic institutions on cutting edge research projects, publishes peer-reviewed papers regularly, and is frequently invited to give lectures at major clinical/scientific conferences such as ECR, RSNA, and SIIM. 

Protege’s Key Value For Data Partners & AI Model Developers

Within this joint approach, Protege acts as the glue that assembles diverse data sources — starting with Segmed’s imaging depth — into a single, AI-ready product. This provides the connective tissue between data partner capabilities and the requirements of the end AI model builder purchasing the data, enabling quick data delivery cycles and quicker model training. 

In the AI training and evaluation world, Protege brings those pieces together into a unified, standardized package so that AI developers get exactly what they’ve been asking for: an integrated, enterprise-ready multimodal dataset that accelerates real model progress.

However, that process isn’t necessarily straightforward. In many cases, AI model builders are pushing the frontier of how foundational AI models perform, and in some cases, are attempting use cases that have not previously been used in clinical or commercial scenarios.

That’s where Protege’s Data Lab expertise comes in — by providing AI-data specific expertise around dataset curation, cleaning, pruning, and delivery, the team of researchers and scientists helps model builders transform datasets provided by some of the world’s largest healthcare data providers like Segmed into AI training and evaluation-ready datasets and products that help improve model performance and outcomes.

For data partners, Protege focuses on turning existing assets into repeatable programs rather than one-off exports. That includes:

• A clear revenue model tied to curated, deal-driven packages

• Shared compliance and governance responsibilities, reducing the burden of supporting complex AI buyers alone

• Co-marketing that increases repeat deal and expansion opportunities as successful programs expand

For AI developers and model builders, Protege concentrates on multimodal performance and speed to deliver one standardized package with unified governance and documentation, combined with delivery patterns that map cleanly to how foundation model teams expect to consume data. 

Unlocking New AI Advancement with Protege

Medical data companies like Segmed partner with Protege to create AI developer-ready training and evaluation datasets that are built upon their existing data assets. 

Protege helps deliver custom curated datasets to the largest foundation AI model builders across the globe. This unlocks new revenue opportunities for data partners large and small that typically would require a full AI-specific business model. 

To learn more about how your healthcare organization can license their data assets to AI companies through Protege, or how your healthcare AI organization can get access to curated AI training and evaluation datasets, contact our healthcare team at contact@withprotege.ai.