Digital Endpoint Data Scientist

Turning Digital
into Clinical

I work with clinical teams to turn sensor signals into endpoints that regulators trust. Less buzzwords, more evidence that holds up when it matters.

See What I've Built Say Hello
6+ Years in
Digital Health
PhD Wearable Health
Technology
10+ Clinical Teams
Supported
Domenico Minici

About

I'm a Sicilian who ended up in Basel doing something I genuinely love: figuring out how to make wearable sensors useful for clinical research.

At Novartis, I work with clinical, regulatory, biostatistics, and operations teams to design digital endpoint strategies — not just technically, but in real trials with real patients. That means thinking about validation, usability, and all the things that can go wrong.

My PhD was about detecting frailty in older adults using wrist-worn accelerometers. I spent years learning how messy real-world sensor data can be, and how to extract something meaningful from it. That experience shapes how I approach every project now.

Outside of work, I'm pretty simple: family,morning runs or time at the gym, dinners with friends. I believe good work happens when people enjoy working together, so I try to bring that energy to my teams.

Basel, Switzerland
PhD in Smart Computing
Open to: Switzerland & Slovenia
Download CV

Where I've Been

Apr 2023 – Present

Digital Endpoint Data Scientist

Novartis Pharma AG • Basel, Switzerland

Working across therapeutic areas to make digital endpoints a reality in clinical trials. Building toolkits, running feasibility studies, and connecting the dots between tech vendors, clinical teams, and regulators.

Jan 2022 – Mar 2023

Head of International Sales & Data Innovation

Prometeo BV • Rotterdam, Netherlands

A detour into the business side. Built a data platform from scratch, learned a lot about negotiation and cross-border collaboration. Great experience that left me a lot.

Aug 2021 – Nov 2021

Digital Biomarker Data Science Intern

Novartis Pharma AG • Basel, Switzerland

My first taste of pharma. Worked on Alzheimer's research using wearable sensors—got hooked on the potential of digital biomarkers to change how we run trials.

Nov 2019 – Apr 2023

PhD Researcher

University of Pisa • Italy

Four years figuring out how to detect frailty from wrist-worn accelerometers. Built ML pipelines, mentored students, learned that good research means listening to clinicians—not just optimizing metrics.

Education

2019 – 2023

Ph.D. in Smart Computing

University of Florence • Wearable Health Technologies

2016 – 2019

M.Sc. Computer Engineering

University of Pisa

2012 – 2016

B.Sc. Computer & Automatic Engineering

Sapienza University of Rome

What I Do

From picking the right device to proving the endpoint works—I've been through the whole cycle

Picking the Tech

Which device fits this study? Is it ready for primetime?

Testing It Out

Running feasibility studies. Will patients actually wear it?

Watching the Data

Monitoring quality in real-time. Catching problems early.

Proving It Works

Building evidence that this endpoint measures what it claims.

Building Tools

Reusable code so we don't reinvent the wheel every time.

Explaining It

Making sure everyone gets what we're doing and why.

Advanced Analytics & Engineering

  • Python ML pipelines, signal processing
  • R Endpoint analytics, statistical modeling
  • SQL Data extraction, transformation
  • AWS Cloud infrastructure
  • Git Version control, reproducibility
  • AI/ML Automation and machine learning

Digital Endpoint & Clinical

  • DHT Validation Verification, analytical & clinical validation, usability
  • Clinical Data Mgmt GxP, CDISC SDTM & ADaM
  • RWD/RWE Real-world data & evidence
  • Wearable Sensors Accelerometers, physiological signals
  • Therapeutic Areas Neuro, immunology, oncology

Strategy & Operations

  • Endpoint Strategy Fit-for-purpose, decision-grade measures
  • Project Management Evidence generation planning
  • Stakeholder Alignment Clinical, Biometrics, Regulatory, Vendors
  • Capability Building Frameworks, toolkits, governance
  • Communication Translating complexity simply

Things I've Built

A few projects I'm proud of. Everything here is shareable—no confidential details.

Internal Capability

Digital Endpoint Evidence Generation Toolkit

An R-based analytical framework for standardizing validation evidence generation across digital endpoint programs. Designed to scale from single-study analyses to enterprise-wide reuse.

Challenge

Digital endpoint validation requires consistent statistical analyses (known-groups validity, convergent/divergent validity, anchor-based responsiveness, sensitivity-to-change) across studies. Without standardization, each analysis is reinvented, risking inconsistency and slowing timelines.

Approach
  • Designed modular R functions aligned to modern DHT validation thinking
  • Built reproducible report templates (parameterized RMarkdown)
  • Conceptualized Shiny interface for non-programmers
  • Defined "valid dataset" specifications upstream of analysis
Impact

Reduced analysis setup time, improved cross-study consistency, enabled knowledge transfer to new team members. Positioned team for future regulatory submissions requiring robust validation evidence.

R Shiny Validation Reproducibility Internal Tooling
View Case Study
PhD Research

Frailty Digital Biomarker from Wearable Sensors

Developed sensor-derived frailty classification methods using accelerometer data, time-frequency transformations, and machine learning—from signal processing to model validation.

Challenge

Frailty is traditionally assessed via clinical scales, but these are episodic and subjective. Wearable sensors offer continuous, objective measurement—if the signal can be translated into clinically meaningful markers.

Approach
  • Designed data collection protocol for older adults with wrist-worn accelerometers
  • Applied Continuous Wavelet Transform (CWT) to generate time-frequency scalograms
  • Trained shallow ML models and deep learning classifiers (CNNs on scalogram images)
  • Evaluated free-living segmentation strategies and wearable-specific constraints
Impact

Demonstrated feasibility of sensor-derived frailty markers. Published findings contributed to the growing body of evidence for digital biomarkers in geriatric research.

Python Deep Learning Signal Processing CWT Wearables
View Case Study
Process Design

Feasibility & Usability Framework for Wearables in Trials

Designed and led healthy volunteer feasibility studies to evaluate digital health technologies prior to deployment in interventional trials.

Challenge

Deploying a new DHT in a pivotal trial without feasibility data risks high dropout, poor compliance, and unusable data. Feasibility studies de-risk this—but require clear protocols and success criteria.

Approach
  • Defined standardized feasibility metrics: wear time, compliance rate, data completeness, user-reported burden
  • Designed study protocols for healthy volunteer evaluations
  • Acted as lead investigator for multiple DHT assessments
  • Created reporting templates for cross-functional decision-making
Impact

Enabled informed go/no-go decisions for DHT deployment, reduced risk in pivotal programs, and built internal knowledge base for future technology selection.

Feasibility Usability Protocol Design Lead Investigator
View Case Study
Analytics

Digital Health Data Quality & Monitoring

Built data quality monitoring approaches for ongoing digital health studies, enabling proactive intervention when adherence or data integrity issues arise.

Challenge

DHT data is only valuable if participants wear devices as instructed and data flows correctly. Without monitoring, problems surface too late to correct.

Approach
  • Defined key quality indicators: missingness patterns, adherence rates, valid wear windows
  • Built dashboards surfacing actionable insights for study teams
  • Designed QA-friendly outputs compatible with clinical data management processes
  • Established escalation thresholds and remediation workflows
Impact

Enabled early detection of compliance issues, supported site-level interventions, and improved overall data quality in supported studies.

R Data Quality Monitoring Dashboards
View Case Study

The PhD Years

I spent four years trying to answer one question: can a simple wristband tell us if someone is becoming frail?

Assessment of Frailty using a Wrist-worn Device

Frailty in older adults is usually assessed with clinical questionnaires—subjective, infrequent, easy to miss. I wanted to see if we could do better with continuous sensor data.

  • The signal problem: Raw accelerometer data is noisy. I used wavelet transforms to turn it into something a model could learn from.
  • ML vs. deep learning: Tested both traditional algorithms and CNNs. Learned when each one makes sense (spoiler: fancier isn't always better).
  • Real-world messiness: Lab data is clean. Home data is chaos. Figuring out how to segment and filter real-world signals was half the battle.
  • Working with clinicians: The best insights came from geriatricians who actually treat these patients. Research in isolation doesn't work.
What I Used
Accelerometry CWT / Scalograms Random Forest CNN Python scikit-learn TensorFlow
Read the Thesis Google Scholar
Selected Publications

Full publication list available on Google Scholar.

Let's Connect

Whether you're working on digital endpoints, thinking about a career move, or just want to chat about wearables and data—I'm always happy to talk.

I respond to everyone. Coffee chats are welcome if you're ever in Basel.

Based in

Basel, Switzerland
Open to roles in Switzerland & Germany

LinkedIn

Let's connect

Grab My CV
Say Hello