Caroline Flessa, MPH | Clinical Data & Digital Health Consultant

0+
Years across clinical and adjacent research: <0 mo
To first in-house wearable QC framework: 0%
On-time SAS exports at Clario: 0+
Concurrent global studies led

I'm a clinical data professional with 10+ years across clinical and adjacent research, spanning academic research, pharma, CROs, and non-profits. My work has concentrated in digital health technology, digital measures, and sensor-derived data.

At Pfizer, I directed the design and build of the company's first in-house wearable data QC framework, with no existing model to build from. I independently delivered the full documentation suite: modular Python QC code, a metrics framework and specifications, validation logs, dataflow maps, and project timelines, and I co-developed the master technical QC requirements with the internal data science team. The framework was handed over complete and ready for testing. I'm also experienced in preparing and delivering clean eCOA datasets and documenting data management approaches for regulatory and internal governance.

I hold an MPH in Epidemiology and Biostatistics, with graduate and post-graduate training in reliability and validity. I work in Python and have applied Databricks in coursework; I actively take on AI and data engineering training as the tooling in this field evolves. My therapeutic area experience covers oncology, dermatology, rheumatology, immunology, respiratory, reproductive endocrinology, and CNS, with CDM experience across Phase I–III.

OncologyDermatologyRheumatologyImmunologyRespiratoryReproductive EndocrinologyCNS

Career Path

Associate Director, Data Operations, Digital Sciences · Atrium, embedded at Pfizer

Nov 2024–Apr 2025

Directed design and build of first-of-its-kind wearable data QC framework, delivering five deliverable categories in under six months. Partnered with data science, biostatistics, and clinical operations as Sponsor point of contact; co-developed master technical requirements and standardized device QC for re-use across studies. Structured documentation suite for regulatory and internal governance, and audit readiness.

PythonSensor Data QCValidation LogicAudit Readiness

Education & Certifications

MPH, Epidemiology & Biostatistics
Boston University School of Public Health
BA, Psychology, Cum Laude
Framingham State University
Clinical Trials Operations Specialization, Johns Hopkins (2026) | ICH GCP, TransCelerate & ICON (2025) | Veeva Vault CDMS & CDB | REDCap Cloud
Clinical Operations & Compliance
Databricks (Generative AI, AI Agents, Data Engineering) | Neo4j | Anthropic Model Context Protocol (2026)
AI & Data Engineering
Responsible AI Governance, 5-course series, All Tech Is Human | Ethical Data Handling, Fred Hutch (2026)
Responsible AI & Data Governance
Genentech Data Sciences in Pharma (Real-World Data, Patient-Centered Outcomes, HTA) | Making Data Science Work for Clinical Reporting, Genentech (2026) | ML in Healthcare, Cleveland Clinic | PySpark, Biopython, RAG (2026)
Data Science in Pharma & Health
Foundations of Agile Project Management, Google (2026)
Project Management

Community

Women Applying AI (WAAI): Web Content Manager (2026)
Healthcare Businesswomen's Association (HBA): Connector (2026); Mentee (2025); presented on Executive Presence & Personal Branding to cohort as part of Mentee Program (2026)

Endorsements

“Caroline's blend of technical expertise, attention to detail, and collaborative ethos make her an exemplary candidate for any Clinical Data Manager role. I am confident that she would bring a wealth of knowledge and a positively proactive approach to any team she joins.”

Karnaditya Rana, MBA, MPHAssociate Director, Clinical Data Management · Multi-omics, Precision Medicine & Cancer Gene Therapy

“Caroline would be an asset for any data or project management roles requiring leadership, critical and analytical thinking, and advanced technical knowledge. What stood out most was her dedication to producing quality work, her ability to collaborate, and her consistency in meeting deliverables and deadlines.”

Jessica AbramsDevelopment Director · Kupenda for the Children

Let's Connect

Open to full-time clinical data and data operations roles. If you're building a team that works with digital measures or sensor data, let's talk.

Get in touch →cmflessa@proton.me

Design for reuse from the start

Modular code, standardized specs, and documentation built to last, so a QC framework or data system can be picked up and applied across studies, not rebuilt each time.

Document for the auditor

Every workflow I build is structured for regulatory and internal governance from day one: validation logs, dataflow maps, and inspection readiness built in, not added at the end.

Work across the aisle

I've worked directly with data science, biostatistics, and clinical operations teams to turn ambiguous requirements into shared technical specs, and to keep data decisions traceable back to a clinical rationale.

Know the tooling, know its limits

I actively pursue Python, Databricks, and AI coursework so the data practices I recommend reflect where the field is going, and so I can tell the difference between a useful tool and an unvalidated one.

Selected Work · tap a row to expand

Challenge

A multi-site wearable study needed a QC framework for continuous sensor data with no in-house process or vendor precedent to build from.

Approach

Directed the design and build, independently delivering five deliverable categories: modular Python QC code, a metrics framework and specifications, validation logs, dataflow maps, and project timelines, and co-developing the master technical QC requirements with the internal data science team.

Outcome

First-of-its-kind wearable QC framework for the organization, designed for reuse. Complete and ready for testing on handoff; documentation structured for inspection readiness and regulatory governance throughout.

Challenge

A high-visibility global portfolio needed eCOA implementations stood up quickly and correctly, with only one US-based CDM available to lead.

Approach

Led study startup as the sole US-based CDM in charge of implementations, authoring core deliverables (DMPs, DCPs, edit check specs, database build files, and data transfer documentation) while implementing proprietary eCOA SaaS to Sponsor protocols.

Outcome

Led startup across 10+ concurrent global studies, was retained post-handover as the subject matter expert on amendment and data transfer impacts, and trained 3–5 CDMs during a period of restructuring.

Challenge

15+ concurrent global studies required test and production SAS data exports delivered on time without compromising data quality.

Approach

Owned eCOA data responsibilities from study setup through database lock, diagnosing and resolving code issues to protect both data quality and timelines.

Outcome

Achieved 100% on-time test and production SAS exports across the portfolio and received the ERT Endeavor Award for measurable, sustained client impact.

Challenge

A digital health startup needed to choose an EDC platform and establish regulatory-aligned governance for sensor and outcome data.

Approach

Advised on data strategy and EDC platform selection, assessing options against regulatory requirements for sensor and outcome data governance.

Outcome

Delivered a regulatory-informed vendor comparison and integration roadmap to the CSO and team within the first month of the engagement.

Challenge

Kupenda needed a mobile Child Case Management app and supporting data systems built to serve children with disabilities in low-income communities in Kenya, with no dedicated technical staff.

Approach

Provided pro-bono consulting through BU School of Public Health, designing the app prototype and aligning content with organizational needs. Extended the engagement a full year post-coursework to collaborate directly with US developers and Kenyan app users on refinement and scale. Generated and validated custom SAS code to analyze piloted app data in-house.

Outcome

Delivered a functioning app scaled to Kilifi, Kenya, an in-house SAS analysis capability, and conference abstracts and poster materials presented at global health conferences.

Data Science & Analytics

PythonReproducible WorkflowsDataset PrepLongitudinal & Observational DataData Summaries & VisualizationDatabricksSASSQL

Digital Measures & Endpoints

Wearable & Sensor Data QCDigital BiomarkersMeasurement Validation (reliability, validity)eCOA / ePROEndpoint Data Documentation

Data Quality & Governance

Data Governance & QCValidation Logic & SpecsRBQMAudit & Inspection ReadinessUAT

Standards & Regulatory

CDISC / SDTMGCP / ICH E6(R3)21 CFR Part 11Documentation for Regulatory

Clinical Data Management

Study Startup LeadershipDatabase BuildData Review & CleaningDatabase Lock & CloseoutEDC Systems & ConfigVeeva Vault CDBREDCapOracle

Live Tools & Products Live items link out · Gated items available on request

Live

AI Solution Quality & Risk ScorecardIndependent / Open Source · 2026

AIPythonRegulated IndustriesLive

Live

Trial Terms: Negotiation Prep for Clinical Data & R&D DealsIndependent / Open Source · 2026

StrategyNegotiationResponsible AIClinical OpsLive

Challenge

Providers and health systems have no structured, compliance-first way to discover and compare clinical AI models. Regulatory status, evidence level, and bias disclosures sit scattered across vendors, so model selection happens without governance.

Approach

Designed a clinical decision-support platform with multi-parameter search, side-by-side comparison of up to three tools, and a guided selection wizard that surfaces FDA clearance status, EU MDR/IVDR marking, evidence level, and documented bias and limitations on every tool. NPI-verified provider and institutional access with role-based controls, plus a hybrid admin-seeded and vendor-submission catalog.

Outcome

Gated demo showing compliance-first AI model discovery, comparison, and selection, with institution-level adoption analytics and a clinical audit view by patient complexity. Built on Base44 with row-level security.

Request access to this demo

Challenge

People navigating chronic GI and autoimmune diagnostic limbo lack tools that combine daily symptom tracking, validated peer support, and doctor-ready summaries, leaving them without data or validation between appointments.

Approach

Designed a self-management app around daily pulse check-ins (gut, energy, emotion), food logging with delayed-reaction severity, a community-contributed nutrition database, peer support circles by symptom cluster, and AI-generated plain-language visit summaries. Row-level security scopes personal health data to its owner, and a mentor-alert automation triggers on low scores.

Outcome

Gated demo illustrating patient-centered symptom tracking, peer community, and a printable two-week trend report for clinician visits. Built on Base44.

Request access to this demo

Challenge

Job seekers, especially healthcare workers facing burnout, juggle pipeline tracking, interview prep, and negotiation across scattered tools, none of which blend emotional support with practical career strategy.

Approach

Designed a dual-variant platform (Opus for general professionals, Helix+ for healthcare workers) from a single codebase: a Kanban pipeline, automated Gmail and IMAP sync that builds the pipeline, AI fit scoring, mock interviews, resume tailoring, and wellness-first coaching. Tiered access with row-level security and a documented Terms and Privacy layer (CCPA/GDPR).

Outcome

Gated demo showing end-to-end job-search management with AI coaching, email-driven pipeline automation, and healthcare-specific interview and salary intelligence. Built on Base44.

Request access to this demo

Challenge

AI-assisted AE coding tools can introduce automation bias when the governance layer is absent: reviewers approve proposals without working exceptions, and weak matches or serious-event misclassifications go undetected.

Approach

Browser-based governance console built on a fully fictitious dummy study (VRN-204) with synthetic, non-patient data: AI proposes MedDRA terms with confidence scores; reviewers work exceptions (Approve / Remap with reason / Escalate to medical review). Designed against FDA 2025 AI guidance and 2026 FDA/EMA good AI practice: narrow context of use, risk-matched oversight, full decision provenance.

Outcome

Functional demo on dummy data, not a real study: illustrates compliant AI-assisted AE coding governance, where every decision logs reviewer, timestamp, and reason for audit readiness.

Challenge

Show a measurable physiological response to a calming intervention within a single hackathon day, without overclaiming from what an n=1 design can actually support.

Approach

Led measurement and methodology: designed a within-subject A/B protocol where the participant serves as their own control, moving through baseline, stress induction, intervention, and recovery while capturing HR, HRV (RMSSD), and self-reported distress. Built the honest-caveats framing directly into the deck (n=1, single session, fixed order, known confounds like decay and demand effects) so the result was never presented as more than a proof of concept.

Outcome

Presented a cumulative snapshot of the day's data capture, centered on a single-participant A/B run. The tool was shown live, though the session leaned on talking points and Q&A rather than a guided interactive walkthrough.

Challenge

No reusable framework existed for QC of continuous wearable sensor data across a pharma portfolio.

Approach

Directed the design and build, independently delivering five deliverable categories: modular Python QC code, a metrics framework and specifications, validation logs, dataflow maps, and project timelines. Master technical QC requirements co-developed with the internal data science team.

Outcome

Full framework complete and ready for testing on handoff; designed for reuse and structured for inspection readiness throughout.

Deployed in Kenya

Child Case Management Mobile AppKupenda for the Children · 2020–2021

mHealthUX DesignSASLive

Live

WAAI Website & Content InfrastructureWomen Applying AI (WAAI) · 2026

Web ContentAI CommunityManagementLive