PROJECTS & INNOVATION

Projects & Innovation

Translating scientific research, clinical insights, and emerging technologies into evidence-driven solutions.

24,167Oncology patients analyzed
200,513Authorization records processed
0.97Mean clinical evidence extraction F1
0.88Retrospective denial-risk ROC-AUC

01 / HEALTHCARE AI & CLINICAL WORKFLOW INNOVATION

Hidalga Technologies

Translating oncology workflow problems and clinician insights into a responsible AI research and product roadmap.

Full product leadership case study →
OVERVIEW

Founded a healthcare AI venture investigating how clinical evidence readiness, payer-policy requirements, and administrative workflow design can improve prior authorization preparation in oncology.

THE PROBLEM

Oncology practices face fragmented payer rules, missing clinical evidence, disconnected systems, and repeated administrative review. Better decisions require understanding the human workflow, not simply automating forms.

MY ROLE

Founder, CEO, and Principal Investigator. Personally owned the product roadmap, prioritized features, defined requirements, supervised two employees and ten contractors, led NSF-funded R&D, and contributed hands-on research and model development.

APPROACH

Translated discovery findings across more than 340 documented interviews in the broader program into priorities for documentation readiness, evidence extraction, payer-policy retrieval, denial-risk analysis, and human review. Established traceability and uncertainty-handling requirements.

IMPLICATIONS

Connected clinical stakeholder evidence to technical requirements and a prospective validation strategy. These are research and product-development outcomes, not demonstrated production deployment or realized clinic savings.

02 / CLINICAL DATA SCIENCE & PREDICTIVE MODELING

Oncology AI Research Pipeline

Reproducible clinical NLP, predictive modeling, longitudinal data analysis, and defensible retrospective evaluation.

Full clinical data science case study →
OVERVIEW

Personally developed a 39-stage Python research pipeline for longitudinal oncology EHR exports, integrating cohort construction, clinical information extraction, classification, model validation, and reproducible reporting.

THE PROBLEM

Retrospective authorization data can be incomplete, imbalanced, and sensitive to time leakage. Extracting clinical evidence and investigating denial risk require careful outcome definitions, patient grouping, and quality controls.

MY ROLE

Hands-on data scientist and primary developer of the retrospective pipeline, feature engineering, evidence extraction, and denial-risk evaluation.

APPROACH

Harmonized heterogeneous clinical records, analyzed missingness and provenance, developed structured and text-based extraction, and evaluated rare-event prediction with patient-grouped splits, PR-AUC, ROC-AUC, and temporal checks.

IMPLICATIONS

Demonstrated feasibility of evidence extraction and retrospective risk modeling while exposing the need for prospective instrumentation and real-workflow validation.

03 / COGNITIVE NEUROSCIENCE & EXPERIMENTAL RESEARCH

Smartphone Notifications & Cognitive Control

Experimental investigations of how digital interruption influences behavior and neural measures of attention.

Full neuroscience case study →
OVERVIEW

Designed and conducted human neuroscience experiments examining the behavioral and electrophysiological effects of smartphone notifications during attention-demanding tasks.

THE PROBLEM

Digital notifications compete for attention, but behavioral measurements alone do not explain the neural processes involved or how interventions may affect them.

MY ROLE

First-author researcher responsible for experimental design, participant-study execution, EEG analysis, statistical interpretation, and manuscript preparation.

APPROACH

Used visual oddball paradigms, event-related potentials, spectral EEG, behavioral performance, and a randomized mindfulness/control induction to investigate cognitive control under distraction.

IMPLICATIONS

Produced peer-reviewed evidence relating technological distraction and neural markers of attentional control, with potential applications in human-centered device and interface design.

04 / EYE TRACKING & EMOTIONAL ATTENTION

Doctoral Dissertation: Emotional Attention & Eye Movements

A preregistered study of emotional facial cues, smartphone notifications, and attentional orientation.

Explore doctoral research →
OVERVIEW

Completed doctoral research examining how emotional faces and smartphone notifications jointly influence visual attention, behavioral performance, and physiological eye-movement measures.

THE PROBLEM

Emotional salience and digital alerts can interact in ways that are not captured by response time alone. Eye tracking allows a more detailed view of overt attentional shifts.

MY ROLE

Doctoral researcher responsible for preregistration, experimental design, implementation, participant sessions, eye-tracking analysis, statistical modeling, and dissertation interpretation.

APPROACH

Used a 360-trial dot-probe task with emotional face stimuli, auditory notifications, and high-frequency eye tracking to study fixations, saccades, pupil responses, and individual differences.

IMPLICATIONS

Connected fine-grained psychophysiological measures with questions of technology-related distraction and affective attention.

EXPERIMENTS & PROTOTYPES

Additional Projects

Selected technical implementations supporting research, feasibility evaluation, and product decisions.

01

APPLIED AI

Local Payer-Policy RAG System

Built a Python/FastAPI retrieval prototype processing 68 payer-policy PDFs into a 1,753-entry FAISS index with grounded local Llama-based question answering.

Technical portfolio →
02

DATA VISUALIZATION

Streamlit Financial Modeling Application

Developed an interactive scenario-analysis tool for projected authorization workload, administrative costs, avoided denials, and modeled return on investment, with Excel outputs.

Relevant case study →
03

AI AGENTS

Azure AI Agent Prototypes

Implemented an Azure AI Foundry agent connected to a seven-node BuildShip workflow for generating synthetic FHIR R4 bundles, with response synchronization undergoing testing.

Explore capabilities →

RESEARCH & DEVELOPMENT

Interested in the evidence behind the work?

Explore research methods, model results, limitations, and selected public-safe artifacts in my technical portfolio.