Author

Massimo

Speaker at the international conference “Innovation Roundtable”, 22 March 2022

by Massimo 2022-02-08

It is a honor for me to speak at the next conference of Innovation Roundtable in Lyon on the 22nd March 2022.

Link to the website of the conference: www.innovationroundtable.com

Title: DSM • Organizing and Managing AI and Data-Driven Projects – How to Set Up Digital Hubs, Drive Change, and Manage Challenges in AI Projects – Massimo Buonaiuto, EMEA Principal Scientist

Abstract

Massimo Buonaiuto will talk about how to organize for and manage AI and data-driven projects. He will explain how organizations can set up digital hubs depending on their strategy and existing infrastructure. Massimo will also discuss lessons learned including key elements to ensure change management success, as well as explain how DSM set up projects and organizational structures. He will also provide use cases to illustrate how to manage AI projects and deal with challenges.

How to structure the company to face the challenges of data science:

- Designing digital hubs – different ways of setting up digital hubs depending on existing infrastructure and long-term strategy
- Identifying areas of impact of digital transformation – and how to prioritize depending on impact
- Change management practices – the importance of communication, transforming teams into champions of change, and changing resource allocation
- How to manage the lack of best practices and avoid expensive fails in AI

Use cases and managing AI and data-driven projects:

- How DSM set up projects and implemented organizational structures to ensure success
- How to deliver the answers that data science aims to provide – and manage high levels of uncertainty
- How to deal with the AI black box – how to explain the results of AI models and ensure the value of prediction
- How to set up data governance and align it with business needs

AGENDA OF THE CONFERENCE

2022-02-08

Artificial Intelligence Data Science and Governance Neural Network

Interleaving algorithm for optimization of neural networks with self-learning perceptrons

by Massimo 2022-01-01

Interleaving is an increasingly popular technique for evaluating information retrieval systems based on implicit user feedback

In one of my research project on data science, I coordinated a new effort to select the best compund from a list of compounds using the interleaving approach, instead of the common A/B Testing,

In particular I used the team draft interleaving, which mimics the process of how team selection occurs for a friendly sports match.

You can find more about this in these links:

https://dl.acm.org/doi/10.1145/2094072.2094078

https://netflixtechblog.com/interleaving-in-online-experiments-at-netflix-a04ee392ec55

This approach can be implemented also in Artificial Intelligence models where perceptrons change their activation functions based on the input data, after a pattern of data is recognized from the perceptron itself or from the control part of the model. In the past, we used standard backpropagation of the error to improve the weights, but we found sometimes perceptrons can change even before the process of learning is completed. This learning during responding has been possible thanking the interleaving approach to choose the best configuration for this model.

2022-01-01

Data Management Data Science and Governance R&D

Unlocking the Power of Data Standardization: A Guide to CDISC and FAIRplus Cookbook for Streamlined Research. Transforming data management to foster adoption of FAIR principles

by Massimo 2022-01-01

Looking for a resource that provides guidance on FAIR data management in Life Sciences? Look no further than the FAIR Cookbook! This resource is a collection of recipes that offer a deep dive into the technical aspects of FAIR data management and the infrastructure needed. It also provides applied examples of FAIRification with clinical trial, epidemiological, and molecular data, making it a valuable resource for data stewards, data managers, and data curators. One of the recipe of this great resource, if the application of FAIR principle in a context of CDISC-SDTM, a standard model for organizing, annotating, and formatting data from clinical trials.

The FAIR Cookbook is a live resource that provides guidance on how to implement the FAIR Principles in the Life Sciences. It is intended for a wide range of audiences, including researchers, data stewards, software developers, policymakers, and trainers. The recipes are a combination of guidance, technical, hands-on, background, and review types, and are classified according to the intended audience. It is based on a grant funded by UE and IMI. It is an international effort that included the Countries listed on the right side.

The FAIR Cookbook is a community-driven resource that is being populated and improved iteratively in an open manner. It is funded by the IMI FAIRplus project, which comprises a coalition of Europe’s leading experts in data interoperability, standards, pre-clinical to clinical translation, and long-term sustainable data repositories. The Cookbook also links to relevant community, mature, and complementary resources in the Life Sciences, such as the RDMkit, FAIRsharing, biotools, and TeSS, as well as the Pistoia Alliance’s FAIR Toolkit for Life Science Industry and more generic resources, such as ‘The Turing Way’ handbook for reproducible data science.

The FAIR principles provide guidelines for making data more easily findable, accessible, interoperable, and reusable. These principles have been widely adopted by funding agencies, industry leaders, and organizations in various fields.

The FAIRplus Cookbook uses these principles as an organizing principle and offers atomic recipes for implementing them.

Ensuring that data is FAIR is important for facilitating data sharing and reuse, which can lead to more efficient and effective scientific research. However, there are also ethical considerations surrounding data sharing and reuse, such as ensuring data privacy and security and obtaining informed consent from study participants. These issues are explored in the FAIRplus Cookbook.

The first time I have been in touch with this precious service was during the mapping of clinical trial data to CDISC (Clinical Data Interchange Standards Consortium).
CDISC-SDTM is a standard model and framework for organizing, annotating, and formatting data from clinical trials. Regulatory agencies such as the US FDA require clinical trial results to be submitted in this format. The SDTM Implementation Guide (SDTMIG) provides expanded guidance on implementing SDTM for specific use cases or “domains”. CDISC standards provide procedures and guidelines for encoding project-specific data that might not fit into the existing domains.
Mapping a non-conforming data dictionary to CDISC STDM standard and CDISC Vocabulary presents some challenges. Given the complexity of CDISC standards, any project team intending to convert their datasets to SDTM at any point in the project life cycle should aim to align with the standard as early on in the process as possible. Data dictionaries and data collection instruments should, where possible, be aligned to the relevant CDISC standards to facilitate data conversions later on. The FAIRcookbook site was very useful to initiate the project for the tabulation model.
CDISC provides a comprehensive framework for the standardization of data collection, organization, and analysis in clinical trials, with the aim of improving data quality, consistency, and interoperability.
The main parts in CDISC (Clinical Data Interchange Standards Consortium) are the following:
– CDASH (Clinical Data Acquisition Standards Harmonization): CDASH provides guidelines for the collection of data during clinical trials, with a focus on data quality and consistency. It aims to standardize the collection of data across multiple trials, making it easier to compare and combine data from different sources.
– SDTM (Study Data Tabulation Model): SDTM is a standard format for organizing and presenting data from clinical trials. It provides a framework for the tabulation of clinical trial data in a consistent and standardized way, making it easier to analyze and compare data from different sources.
– ADaM (Analysis Data Model): ADaM provides a standard format for the analysis of clinical trial data. It provides a framework for the creation of analysis datasets that can be used for statistical analysis, reporting, and submission to regulatory agencies.
– Define-XML: Define-XML is an XML file that provides metadata for the datasets used in a clinical trial. It describes the variables, datasets, and relationships between them, providing a clear and comprehensive overview of the data used in the trial.
– Controlled Terminology: Controlled Terminology provides a standardized set of codes and definitions for data elements used in clinical trials. It aims to ensure consistency and accuracy in the use of terminology across different trials and data sources.

Another part of the CDISC standard I worked on is the SEND, which provides a way to exchange non-clinical data between organizations, such as pharmaceutical companies and regulatory agencies. The SEND Implementation Guide defines the structure and content of non-clinical data that should be submitted to regulatory agencies in order to support the evaluation of a drug’s safety and efficacy. The SEND Implementation Guide specifies the format and content of datasets for non-clinical studies, including general toxicology, reproductive toxicology, carcinogenicity, and other studies. The guide provides detailed instructions for the organization of data, variable naming conventions, controlled terminology, and other elements necessary for standardized data exchange. The SEND Implementation Guide is intended to be used in conjunction with other CDISC standards, including the Study Data Tabulation Model (SDTM), which provides a standard format for clinical trial data.

Together, these standards help ensure that data is submitted in a consistent and standardized format, which facilitates data exchange, analysis, and evaluation.

2022-01-01

About me Public speaking events

Speaker at “8th Artificial Intelligence, Data Analytics and insights summit – DACH Region”, 11th-12th November 2021

by Massimo 2021-11-09

It is a honor for me to speak at the next conference “Speaker at “8th Artificial Intelligence, Data Analytics and insights summit – DACH Region”, 12th November 2021. Together with 250 senior managers and international leaders in Artificial Intelligence, data governance and research in data science, I will discuss about our research in Artificial Intelligence applied to Manufacturing and the challenges behind data governance in the pharma industry.

Link to the website of the conference: https://berryprofessionals.com/ai-data-analytics-and-insights-summit-dach/

AGENDA OF THE CONFERENCE

2021-11-09

Data Management Data Science and Governance

Total Data Quality Management: A Comprehensive Approach to Data Quality

by Massimo 2021-08-15

Total Data Quality Management (TDQM) is a comprehensive approach to data quality that covers every aspect of data management, from data governance and data strategy to data modelling and data analysis.

TDQM focuses on ensuring that data is accurate, complete, consistent, timely, relevant, accessible, usable, transparent, and secure, and that it meets the needs of the business.

Total Data Quality Management is a comprehensive approach to managing data quality that ensures that data is accurate, complete, and consistent. It involves a range of activities, including data profiling and analysis, data cleansing, data governance, and data quality monitoring. TDQM helps organizations to improve their data quality, reduce data-related risks, and make better-informed business decisions. By implementing TDQM, organizations can ensure that their data is a strategic asset that supports their business objectives and enables them to compete in today’s data-driven business environment.

TDQM involves several key components, including data governance, data management, data strategy, data modeling, data analysis, data integration, data security, data privacy, data accuracy, data completeness, data consistency, data timeliness, data relevancy, data accessibility, data usability, data transparency, and data lineage. These components are interrelated and work together to ensure that data quality is managed effectively.

Data Governance

Data governance is the foundation of TDQM. It is the process of defining and implementing policies, procedures, and standards for data management. Data governance involves identifying the stakeholders, roles, responsibilities, and processes for managing data. It also involves defining the rules for data quality and ensuring that they are adhered to.

Data Management

Data management involves the processes and technologies for managing data throughout its lifecycle. This includes data integration, data modeling, data analysis, data quality assurance, data quality control, and data quality improvement. Data management ensures that data is accurate, complete, consistent, timely, relevant, accessible, usable, transparent, and secure.

Data Strategy

Data strategy is the process of defining the goals and objectives for data management. It involves identifying the business needs for data, the data requirements, and the data sources. Data strategy also involves defining the metrics for data quality and measuring the performance of data management.

Data Modeling

Data modeling is the process of defining the structure and relationships of data. It involves creating a data model that represents the data in a way that is understandable to the business. Data modeling ensures that data is consistent and that it meets the business requirements.

Data Analysis

Data analysis is the process of analyzing data to gain insights into the business. It involves using statistical methods and data visualization to identify patterns and trends in the data. Data analysis helps to identify data quality issues and to improve the quality of data.

Data Integration

Data integration is the process of combining data from different sources into a single view of the data. It involves identifying the data sources, mapping the data, and transforming the data to ensure that it is consistent. Data integration helps to ensure that data is accurate and complete.

Data Security and Privacy

Data security and privacy are critical components of TDQM. Data security involves protecting data from unauthorized access, use, disclosure, or destruction. Data privacy involves protecting personal information from unauthorized access or use. Data security and privacy are essential for maintaining the trust of customers and stakeholders.

Data Profiling and Analysis
Data profiling and analysis is an essential component of TDQM that involves the examination of data to understand its structure, content, and relationships. Data profiling is a process of gathering and analyzing metadata to understand the characteristics and quality of data. It provides a better understanding of the data to identify quality issues and helps to create a baseline for data quality improvement efforts. Data analysis, on the other hand, involves the use of statistical methods to detect patterns, trends, and anomalies in the data. The results of data analysis help to identify root causes of data quality problems, develop appropriate solutions, and monitor the effectiveness of the implemented solutions.

Data Cleansing
Data cleansing, also known as data scrubbing, is the process of detecting and correcting or removing corrupt or inaccurate data from a database. It is a critical step in TDQM that ensures that data is accurate, consistent, and complete. Data cleansing involves identifying and correcting errors such as misspellings, duplicates, and inconsistencies in the data. The process of data cleansing may involve manual or automated methods, depending on the nature and complexity of the data.

Data Governance
Data governance is the process of managing the availability, usability, integrity, and security of the data used in an organization. It involves the development of policies, procedures, and standards for data management, as well as the establishment of roles, responsibilities, and processes for data quality management. Data governance ensures that data is managed consistently across the organization, and that data quality is monitored and improved continuously.

Data Quality Monitoring
Data quality monitoring is the process of continuously monitoring and measuring the quality of data to ensure that it meets the required standards. It involves the use of metrics, indicators, and key performance indicators (KPIs) to measure the quality of data and identify areas for improvement. Data quality monitoring helps to ensure that data quality is maintained over time, and that data is fit for the intended purpose.

2021-08-15

About me Public speaking events

Speaker & Chair at “Pharma Digital Transformation Conference”, 25th November 2021

by Massimo 2021-07-24

It is a honor for me to speak at the next conference “Pharma Digital Transformation Conference”, 25th November 2021.

Together with 20 digital leaders, I will discuss about the challenges of digital and data science in the pharma industry.

With Andreas Panteli, Director Digital Integrated Solutions @ Novartis, I will co-chair the opening session of the conference.

Link to the website of the conference: https://pharmadigitalconference.com/

My video on the conference (Link: https://bit.ly/3kq1G5P )

AGENDA OF THE CONFERENCE

2021-07-24

Artificial Intelligence Data Science and Governance Neural Network

The (Un)reliability of Saliency methods – Google Research

by Massimo 2021-07-16

Feature importance evaluation is a fundamental problem in deep model interpretation and feature visualization. Saliency methods is one of the most popular feature importance evaluation methods, assign to each input feature an importance score, which measures the usefulness of that feature for the performance. A large importance score of a feature means a large performance drop when the feature is unavailable. One way to explain deep learning behaviour is the saliency methods. In Google Research page, it is clearly demonstrated that this method does not work.

Indeed, saliency methods aim to explain the predictions of deep neural networks. These methods lack reliability when the explanation is sensitive to factors that do not contribute to the model prediction.
Saliency methods that do not satisfy input invariance result in misleading attribution.

Source: The (Un)reliability of Saliency methods – Google Research

2021-07-16

About me Artificial Intelligence Data Management Data Science and Governance Public speaking events R&D

Speaker at “AI, Data Analytics & Insights Summit – DACH”, 11th – 12th November 2021

by Massimo 2021-07-06

It is a honor for me to speak at the next conference “AI, Data Analytics
& Insights Summit – DACH”, 11th – 12th November 2021. I will meet 250 Senior Participants from DACH Region.
I will talk about Artificial Intelligence applied to Research and Development.

Ai, Data Analytics

& Insights Summit DACH Online

11th – 12th November 2021

250 Senior Participants from DACH Region

Interactive Senior Level Online Meeting

AGENDA OF THE CONFERENCE

2021-07-06

About me Artificial Intelligence Data Management Data Science and Governance

What I learnt talking to people and speaking at Conferences

by Massimo 2021-07-02

Give your audience only what they need to know

When I have worked with people, they tell them too much. That’s the big thing.
They get stuck in the weeds.
The first thing you want to know is who’s in front of me, what do they care about? And how do they like to receive information. And then you tailor your message to their style and less is more.

And here’s why. If they want åmore information, they’ll ask for it. So, give them what they need to know
not everything you know. And I am going to repeat that because it’s so important.
Give them what they need to know, not everything you know. And when you are crisp, they actually retain more. Keep it simple, keep it short.

It’s hard to go in cold and not know someone, but let’s say you do. Then you want to engage them
in a short conversation, ask them some questions. What are you tackling with? What’s important right now? Start with what’s important to them. And I have a whole process that I put people through in how to structure and organize your talk and your message. I call it listener-centered communication.
And what most people do, the big mistake. One of the big mistakes is they’re speaker-entered
not listener-centered. So they start with what’s important to them: “Good morning, today I want to talk to you about my idea”…. They don’t care about your idea! They care about their own self-interest. So lead with what’s important to them. So if you’re talking to a manager and you want to get an extra person on board to help you out, don’t start with I’m overworked and I need somebody to help me.
Start with “I have a way we can be more productive in this department.”. That’s a hook, a grabber. And so when you lead with that, now you have the listener’s attentionbecause what do managers care about? They care about productivity.
Now that you have that person’s attention you can lead them down the path of how you came to that,
what the problem is and how you have a solution. And you can do it in a really short period of time.
I’ve had people use this process and create a whole presentation and deliver it in six minutes.
And it is powerful. so I would say less is more! … but you have to be able to speak the person’s language. So know yourself, know your audience, know your message.

2021-07-02

Large Language Models LLMs and Natural Language Processing (NLP)R&D

Leveraging NLP in Knowledge Management: a Case Study of Lab Document Management

by Massimo 2021-07-01

Knowledge management in complex environments, such as laboratories, requires innovative approaches to handle vast amounts of data efficiently. The advent of Natural Language Processing (NLP) and large language models offers a transformative solution.

This article explores a case study where advanced NLP techniques were applied to manage millions of lab scanned documents.

Project Overview

The project entailed developing a system to manage a substantial repository of scanned laboratory documents. The primary objective was to enhance accessibility and categorization of these documents using state-of-the-art NLP techniques and machine learning models.

NLP Techniques and Models Applied

Topic Modeling & Document Clustering: Utilized for categorizing documents into coherent groups based on their content. This approach facilitated easier retrieval and analysis of documents based on subject matter.
Semantic Similarity Analysis: Implemented to understand the context and deeper meaning within the documents. This technique helped in linking related documents and provided a more nuanced search capability.
BERT and GPT for Keyword Extraction & Synonym Generation: The use of BERT (Bidirectional Encoder Representations from Transformers) and GPT (Generative Pre-trained Transformer) models significantly enhanced keyword extraction. This led to the generation of relevant synonyms, thereby improving document categorization.

Development of Sophisticated Ontology

A critical aspect of the project was the development of a sophisticated ontology. This ontology served as a structured framework of knowledge, representing concepts within the laboratory domain and the relationships between them.

Integration with Public Data Sources: The ontology was integrated with various public data sources. This integration enriched the ontology with external knowledge, making the internal repository more comprehensive.
Improved Data Interoperability: By aligning the internal data structure with external sources, data interoperability was significantly enhanced. This alignment facilitated seamless data exchange and integration, enabling more collaborative and efficient research.
Research Collaboration: The integrated ontology fostered research collaboration. Researchers could easily connect their work with existing knowledge and collaborate based on shared terminologies and concepts.

Outcomes and Impacts

The implementation of advanced NLP techniques and large language models revolutionized the management of lab documents. The enhanced categorization and retrieval system led to a more efficient research process. Researchers could now easily access relevant documents, draw connections, and collaborate more effectively.

The integration of a well-structured ontology with public data sources further amplified the benefits, establishing a more cohesive and collaborative research environment. This approach not only streamlined internal processes but also positioned the organization at the forefront of research collaboration and knowledge sharing.

Conclusion

This case study underscores the potential of NLP and machine learning in transforming knowledge management, particularly in complex and data-intensive environments. By harnessing the power of advanced algorithms and models like BERT and GPT, organizations can effectively manage vast repositories of information, enhancing both internal efficiencies and external collaboration. The success of this project serves as a benchmark for future endeavors in knowledge management and NLP applications

2021-07-01

Artificial Intelligence Data Science and Governance

How many fields in data science?

by Massimo 2021-06-27

Data Science is a research activity mostly

Data-driven scientific discovery is regarded as the fourth science paradigm

The twenty-first century has ushered in a new age that is coined as data science and big data analytics. Data-driven scientific discovery is regarded as the fourth science paradigm. Data science has been a core driver of the new-generation science, technologies and economy, and is driving new researches, innovation, profession, applications and education across both disciplines and business domains.

There are many scientific and technical challenges associated with big data, ranging from data capture, creation, storage, search, sharing, modeling, representation, analysis, learning, visualization, explanation, and decision making. Among the many data characteristics and complexities to be addressed, I mention the hybridization of heterogeneous, multisource, hierarchical,
interactive, dynamic, multidimensional, and quality-poor data mixed with real-time business operations, strategic planning, decision-making, value creation, and future developments.
The field of data sciences and big data analytics have been evolving from statistics since half century ago to broad areas including but not limited to data and signal analytics, knowledge discovery, information retrieval, machine learning, statistics, optimization, computing, and data management. The literature defined areas of Data Science that requires in depth knowledge and pure research to be effective. By synergizing the three big areas—statistics, informatics and computing, data science has been spreading to essential and specific areas such as

data intelligence and complexity analysis
representation, modeling, analytics, mining and learning including statistical and deep learning
computational intelligence including neural networks, evolutionary computing, fuzzy systems
neuroscience and linguistics
behavioral science and social and economic computing
uncertainty and optimization
system and modeling infrastructures and architectures
networking and interoperation
social issues including
privacy, security, trust, value and impact,
enterprises, services, applications, solutions and systems
simulation, visualization and explanation

2021-06-27

About me Artificial Intelligence Data Management Data Science and Governance Public speaking events R&D

Speaker at FUTURE Labs 2021

by Massimo 2021-05-25

It is a honor for me to speak at the next conference FUTURE Labs 2021. I will talk about Artificial Intelligence applied to Research and Development.

What is Future Labs Live?

This is the world’s most diverse, stimulating, and exciting event for the future of all labs.

Start-ups, disruptors and innovators from academia, Biotech, Pharma, Chemicals, Food & Beverage, Agricultural, Materials, FMGC, Architects, Planners and many more are invited to help pave the way and share their experiences and vision for the lab of the future.

Featuring industry leaders sharing their views and expertise across 9 themes including: AI & Machine Learning, Digital Transformation, Lab Operations & Efficiency, Data Management and more. All conference sessions will be in English.

2021-05-25