Bridging the Data Readiness Gap for AI in Life Sciences

Many organizations in the life sciences are eager to leverage AI and machine learning for breakthroughs in drug discovery, precision medicine, and process optimization. However, one of the most significant barriers is the state of their data—often siloed, unstructured, and spread across disparate systems. 

In this Expert Series, Austin Jarrett, Technical Lead at ZONTAL, shares his perspective on the “data readiness gap” and what it means for organizations trying to scale AI/ML. He discusses the challenges of fragmented data environments, the importance of FAIR principles, and how treating data as a strategic asset lays the foundation for successful AI-driven science.

The Foundational Challenge

To start, could you paint a picture of the current data landscape in a typical R&D organization? From your perspective, how severely does this fragmented and non-standardized data environment impede the progress and potential ROI of their AI/ML initiatives today?

In most R&D organizations today, a vast amount of data is generated on a regular basis from a wide variety of sources. Much of that data ends up scattered across disconnected systems—legacy software, proprietary file formats, shared drives, or even individual researchers’ laptops. As a result, a large portion of valuable data is lost, forgotten, or simply too difficult to access and reuse. 

This fragmentation creates a major challenge for organizations trying to leverage AI and machine learning. Often, different parts of the same experiment are stored in separate systems, making it nearly impossible to get a complete, accurate view of the data. Without a centralized, standardized approach to data management, teams struggle to find and trust the data they need. 

Before AI/ML models can be trained, a huge amount of effort goes into cleaning, organizing, and standardizing the data – a process known as data curation. This includes cleaning the data, aligning it to standardized formats, tagging metadata, managing versions, ensuring security, and making data accessible. When this foundation is missing, AI initiatives are severely limited. Poor data quality leads to unreliable models, reduced scalability, and impeded progress on AI/ML initiatives. 

To unlock the value of AI, organizations need to treat data as a strategic asset – investing in platforms and practices that make data accessible, usable, and trustworthy. 

The Downstream Consequences

Could you elaborate on the downstream consequences of this data fragmentation? For instance, how does it concretely impact the day-to-day work of data scientists and researchers, and what are the risks associated with training AI models on inadequately prepared, or “messy,” data?

Data fragmentation and the lack of proper data management lead to many challenges and frustrations in the day-to-day operations researchers and data scientists. Large amounts of time and effort often spent trying to find or interpret old data or in redoing experiments. 

One of the most common challenges is simply finding the right data. Many organizations report that their teams often struggle to locate past experimental results, especially when the data was generated by someone else or created years ago. Even when they do find it, the data may be incomplete, poorly documented, or incompatible with current tools and systems. 

Researchers often find it faster and easier to re-run an entire experiment than to try to track down and reuse existing data. This often leads to wasted time, duplicated effort, and unnecessary costs – especially when experiments involve expensive materials or equipment. 

For data scientists, the situation is equally frustrating. When data is messy, inconsistent, or lacks proper metadata, it becomes difficult to prepare it for machine learning. Models trained on poorly curated data are more likely to produce unreliable, inaccurate, or biased results, undermining trust and limiting the impact of AI initiatives. 

Fragmented and messy data don’t just slow down innovation – they erode the value of past work and increases the risk of flawed insights. Investing in better data management is a strategic step toward to unlocking the full potential of both human and machine intelligence. 

The Ideal State

AI-Ready Data: To bridge this gap, the concept of creating ‘FAIR’ (Findable, Accessible, Interoperable, Reusable) data is often discussed. Before we dive in, could you explain the foundational principles of what it truly means to make complex scientific data AI/ML-ready?

Making scientific data “AI/ML-ready” goes beyond just collecting it – it requires transforming raw, complex data into a structured, standardized, and enriched format that machine learning models can learn from. 

A key part of this process is data normalization – ensuring that values are consistent, units are standardized, and formats are aligned across datasets. This allows AI models to interpret the data correctly and draw meaningful patterns from features in the data.  

Another important component is metadata enrichment, or adding context to the data, such as where it came from, how it was generated, and what it represents. For example, ZONTAL pulls metadata not only from the file itself but also from additional sources like storage locations, naming conventions, format definitions, or APIs. This added context improves the findability, completeness, and usability of the data, which are essential for both human researchers and AI systems. 

Ultimately, preparing data for AI/ML means making it FAIR: 

• Findable – easily located by both humans and machines 

• Accessible – available through standardized protocols 

• Interoperable – compatible across systems and tools 

• Reusable – well-documented and richly described for future use 

FAIR data is a prerequisite for successful data science efforts, especially when dealing with complex scientific data. 

ZONTAL’s Role

How does ZONTAL position itself to address this challenge? What is its core mission when it comes to transforming scientific data for the AI era?

ZONTAL was built to solve one of the most pressing challenges in scientific research today: transforming fragmented, siloed data into a trusted foundation that drives innovation. 

ZONTAL simplifies and streamlines data management across the entire data lifecycle – from ingestion to long-term reuse – so that scientific data becomes truly AI/ML-ready. ZONTAL serves as a centralized data hub, automatically pulling in data from a wide range of sources, including instruments, file shares, and external APIs. It doesn’t just store the data – it enriches it with metadata, aligns it to standardized schemas and ontologies, and converts data from proprietary formats into vendor-neutral, normalized structures. 

We also ensure that data remains secure, traceable, and compliant. ZONTAL manages version control, access permissions, and audit trails, giving organizations full transparency and governance over their data assets. 

As a cloud-native platform with robust APIs, ZONTAL enables seamless integration with downstream analytics tools and AI workflows. ZONTAL also provides built-in visualization dashboards, allowing users to explore and interact with scientific data directly within the platform – making insights more accessible and actionable.  

 Our goal is to handle the heavy lifting of data preparation, so researchers and data scientists can focus on discovery, not data wrangling. 

 In short, ZONTAL empowers organizations to treat their data not as a byproduct of research, but also as a strategic asset, which is ready to drive scientific discovery and innovation with insights and ROI in the AI era. 

The Harmonization Engine

Could you walk us through how the ZONTAL platform ingests, harmonizes, and contextualizes data from incredibly diverse sources—like different analytical instruments, electronic lab notebooks, and clinical systems—into a single, unified, and analysis-ready format? What makes this process unique?

At the core of ZONTAL’s architecture is a cloud-native, containerized infrastructure that allows the platform to scale dynamically to manage data loads. By leveraging established cloud services, ZONTAL ensures enterprise-grade durability, availability, and security for all ingested data. 

Data can enter the platform through multiple channels, such as automated watchers on files or databases, APIs, or manual uploads. Once ingested, ZONTAL uses flexible data pipelines which leverage an Apache Airflow framework to process and transform the data. These pipelines can be customized to meet the specific needs of different scientific domains or workflows. 

ZONTAL also includes an extensive library of data parsers and connectors which can interpret and extract structured information from a wide range of vendor-specific file formats and data sources. This is critical in scientific environments where data heterogeneity is the norm – different instruments, software versions, and data structures are common even within the same departments and teams. 

Once parsed, the data is standardized and aligned with industry-recognized schemas and ontologies, such as the Allotrope Simple Model (ASM). This harmonization step ensures that data from diverse sources can be understood, compared, and analyzed in a unified way. 

Finally, ZONTAL enriches the data with metadata from both internal and external sources, making it findable, interoperable, and ready for downstream analytics and AI/ML applications. With built-in version control, access management, and audit trails, ZONTAL also ensures full traceability and compliance throughout the data lifecycle. 

From Harmonized Data to Better AI

Once the data is processed by ZONTAL, what does “good” look like from the perspective of an AI/ML algorithm? How does this fully contextualized data directly improve the performance, accuracy, and reliability of predictive models in, for example, a target identification or process optimization workflow?

AI and machine learning models are only as good as the data they’re trained on. If the input data is messy, incomplete, or inconsistent, the models will struggle to learn meaningful patterns – leading to poor predictions, unreliable results, and wasted effort. AI/ML models perform best when they are given a large amount of representative and reliable data to learn from. AI/ML techniques and algorithms can then find patterns in the data through mathematical and statistical methods. 

“Good” data, from an AI/ML perspective, means data that is clean, complete, consistent, and context-rich. It should be well-structured, standardized, and representative of the real-world scenarios the model is meant to understand. This is especially critical in scientific domains where precision and reproducibility are non-negotiable. 

ZONTAL plays a key role in achieving this level of data quality. The platform: 

• Normalizes data across formats and sources, ensuring consistency in units, types, and structure. 

• Validates data early in the pipeline, flagging incomplete or incompatible entries before they reach the model. 

• Enforces schemas and ontologies, so that data adheres to expected formats – for example, ensuring a field meant to contain numerical values doesn’t contain text or symbols. 

• Enriches data with metadata, providing critical context that helps models interpret the information more accurately. 

By delivering normalized and contextualized data, ZONTAL enables AI/ML models to train on high-quality, trustworthy datasets. This leads to: 

• Higher model accuracy, because the data is clean and representative. 

• Improved reliability, thanks to consistent formatting and validation. 

• Faster time to insight, since less time is spent cleaning and wrangling data. 

ZONTAL helps organizations move from “data chaos” to “data confidence” – laying the foundation for AI that delivers real, measurable impact.

The road to AI/ML success in life sciences doesn’t start with algorithms—it starts with data. Without clean, standardized, and accessible information, even the most advanced models fall short. 

As Austin emphasizes, addressing this data readiness gap is essential for accelerating discovery, ensuring compliance, and unlocking the long-term value of research. By harmonizing data and making it FAIR, ZONTAL empowers organizations to shift their focus from data wrangling to meaningful innovation. 

This installment of the Expert Series highlights a key truth: scientific progress in the AI era depends on turning fragmented data into a trusted, AI-ready foundation.