Portfolio

Artificial Intelligence - AI

Data Analytics and Reporting

Designed and launched an AI powered data and analytics platform that integrates Large Language Models and Small Language Models with agentic intelligence, enabling internal and external users to ask natural language questions of EMR data to generate actionable insights. The platform increased customer satisfaction by 65%, reduced time to market from eight weeks to one week, and drove 80% customer adoption. Led product strategy, roadmap definition, and cross functional execution across engineering, data, security, and clinical stakeholders to ensure enterprise scale, compliance, and reliability. Established governance and operating models that enabled rapid iteration while maintaining data integrity, privacy, and regulatory standards. Currently leading the next phase of the platform by integrating language model outputs directly into enterprise BI tools, enabling insights to be operationalized through standardized reporting and executive dashboards.

EMR integrations

Developed and deployed an AI-driven data processing pipelines to automatically ingest healthcare data from multiple EMR sources, including Epic, Cerner, Athena, Meditech, custom SQL databases, and Achadia, and transform it into a standardized format. The system combined NLP with rules-based mapping to extract both structured and unstructured data (e.g., diagnoses, medications, labs, encounter details), normalize key elements (such as codes, units, and timestamps), and align them to FHIR resources, HL7 v2 messages, OMOP, and project-specific custom schemas. This AI-enabled approach significantly reduced manual data cleaning and mapping effort, improved consistency across heterogeneous EMR systems, and created a unified, analytics-ready dataset that could be reliably used for downstream reporting, research, and clinical decision support.

Early Disease & Device Detection

Developed and deployed an AI-based predictive modeling to identify patients at elevated risk for Type 1 diabetes, PCI, MASH, MASLD, Arrhythmias leveraging the standardized data pipeline. Using the harmonized dataset, I trained and validated machine learning models to estimate individual patient risk scores, incorporating both traditional risk factors and patterns not easily captured by rule-based approaches. This AI-driven risk stratification enabled earlier identification of high-risk patients, supported targeted outreach and interventions, and provided explainable outputs that could be reviewed alongside the clinical record to inform care decisions.

Patient Care Pathways

Developed and deployed an AI to model and support the patient care pathway both leading up to and following a diagnosis or device utilization. Using the harmonized data the system analyzed longitudinal patterns in encounters, symptoms, labs, imaging, medications, and clinical notes to identify early warning signals and typical trajectories prior to diagnosis. We were able to distinguish patient care pathways that were directly related to the specific disease or device from those that were unrelated, revealing variation in how patients progressed through the health system. This analysis identified opportunities for more efficient and streamlined pathways leading to diagnosis, as well as the impact that different pre‑diagnosis journeys had on patients’ health status and outcomes post‑diagnosis. Post‑diagnosis, the same framework tracked adherence to evidence‑based care plans, follow‑up visits, key lab and imaging milestones, and treatment response, enabling proactive outreach for gaps in care and more personalized, data‑driven management along the entire care pathway.

Patient Matching

Developed and deployed an AI-powered patient identity resolution engine to accurately match records across fragmented health systems and state registries. We combined supervised machine learning and probabilistic matching techniques to handle data variability, aliasing, and incomplete information. Natural Language Processing (NLP) was used to parse unstructured demographic inputs such as free-text name and address fields, while entity resolution models corrected for misspellings, formatting inconsistencies, and partial matches. This comprehensive approach achieved over 95% match accuracy, significantly improving the continuity of care and compliance for advance care planning (ACP) documentation.

Document Matching Across the US & Document Prioritization

Applied natural language processing (NLP) and classification algorithms to match advance care planning (ACP) documents to patient records across all 50 states. Leveraged large language models (LLMs) to interpret complex legal and clinical language, enabling automated document classification, relevance scoring, and prioritization. AI-powered logic considered document recency, legal status, and state-specific requirements to ensure clinicians saw the most actionable records first. This solution reduced manual review time by over 70% and significantly improved compliance with end-of-life care directives.

Appeal Letters

Leveraged large language models (LLMs) to automatically generate personalized appeal letters that integrate clinical context, patient history, and payor-specific language. This ensured that every appeal was tailored for maximum impact, increasing the likelihood of successful outcomes. By automating this complex process, we significantly reduced full-time equivalent (FTE) overhead dedicated to administrative tasks, while simultaneously improving overall appeal approval rates.

Order Flow Optimizations Based on Payor

Developed machine learning models trained on historical approval and denial data to dynamically adjust diagnostic order workflows in real time. These workflows were customized based on each payor’s policies, coverage criteria, and regional nuances. Through advanced data analysis, we identified the key clinical information required by each payor for first-pass approvals, proactively guiding providers to select the most appropriate test and to include all necessary documentation up front. This approach optimized operations, improved reimbursement success rates, and reduced administrative burden for providers.

Data Validation & Alert Automation from Previous Reports

Applied supervised machine learning and anomaly detection algorithms to validate newly ingested data against historical patterns and known reporting inconsistencies. Enabling automatic flagging of data deviations and contextual anomalies. Built real-time alerting and validation workflows that reduced manual audit effort by over 60% and significantly improved data integrity across quality reporting pipelines.

Additional Projects

Invitae - Genetic Testing

Invitae - Website link

Genetic testing company that provides genetic tests to empower patients the ability to make decisions about their health through the power of genetics.

Product Test Catalog Enhancements

Enhance a test catalog originally designed for Genetic Counselors to be accessible to users without special knowledge of genetics.

Challenge - Enable a user-friendly test catalog that will allow users with varying levels of knowledge in genetics the ability to easily order tests. This was part of a shift-left strategy to expand beyond the limited user-base of GCs (Genetic Counselors - just 2,500 in the United States), and make the catalog more accessible to over 1M US physicians. With these enhancements, just 2-5 orders from users on a monthly basis could expand orders significantly beyond the stagnated 500 a month. Focus is Oncology and woman’s health specialty departments.

Limitations - Current test catalog has technical limitations: some tests we would like to add to the cart simultaneously are not technically possible; due to aggressive growth and acquisitions. There is a plan to address this, but current investment is too large to address at this time. Additionally, current Genetic Counselor workflow should not be impacted with these new workflows as we do not want to impact current order volume.

Vynca - Advance Care Planning

Vynca- Website link

Vynca is an Advance Care Planning Platform focused enabling internal services team members, health systems and health plans the ability to complete advance care planning documents. (Polst/Most, DNR, Advance Directives and Health care Power of Attorney documents) Enabled an e-signature synchronous and asynchronous workflows for one or many document workflows. Vynca is a SaaS and Health Services company focused on Advance Care Planning, Palliative Services and other patient services to provide patient quality of care.

Task Manager

Enable the ability to auto create tasks, assign tasks, auto complete tasks, and allow the ability to connect internal and external users to access the patients records. We also called this a community connect enabling care coordinators the ability to share the status of a patients care plan across their care providers in network and out of network.

Challenge - Build out a more robust platform for internal and external teams to track and document the status of the document workflows for both internally and externally staffed teams. Each document has different workflows and triggers on when to open, send and close appropriate tasks. The focus was to build-out a task management tool that would enable users the ability to manage the patients records and be sure that no patients are left behind.

Limitations - With no technical limitations we had an aggressive push to build as soon as possible. This is a current gap for internal business and impacting team productivity by 25-30%. Current team ~100 FTEs on average $60hr (Excluding 3 Physicians from calculation as hourly rate is significantly higher and we do not anticipate them using this tool directly or any significant time savings. ) $720 per FTE per week $72k for 100 FTEs Impacting $3.7M annually with an expected growth the following year to a team of 250+

Tenant/User Management - Enabled for internal admins to provision tenant users and internal and external users the ability to provision (add, edit & reset password) users permissions.

Tenant Management - Enabled for Internal Admins the ability to add Admin access and turn on/off feature toggles on demand.

Template Builder - Enables the ability to create one or many templates that can be used across tenants. Creating customizable templates for specific document workflows.

Synchronous e-Signature

Challenge - Most healthcare providers do not have touch screens so they need a way to capture a patients and physicians signature during an office visit. An e-signature workflow to enable users to sign via tablet or smart phone while in a live virtual or in person visit of their digital documents.

Limitations - Most healthcare laptops are not touch screen enabled.

Asynchronous e-Signature

Challenge - At times users are not ready to sign while in a visit live and the current offering only offered a live in person or virtual signature. This asynchronous workflow enables the ability to schedule an email/text notification to the signer when their step is ready for review and signatures. (Example: Patient, Witness 1-3 and physician signatures)

Limitations - Health System and Health Plan security requirements to reduce the physician login experience as much as possible. Need to validate customer requirements for 2FA.

Document Upload

Challenge - Enable a quality control for internal and external users to be sure that the correct information is available and that it is an accurate/actionable document. Although the desired output is that all documents are created digitally we found even with the most utilized organizations we still had about a 10-20% paper based workflows needed to support the ability to manually scan and upload a document into the directory.

Limitations - We are still dependent on user error so training is still required for end users. Additionally, each document, field locations and number of questions are different document to document.

Help Guide link

Patient Engagement

Challenge - Enabling Patients the ability to go through a guided digital document completion and review process. Enable patients with a patient portal that can integrate with their EMR Patient Portal experience. The goal is to have a 20% adoption rate to enable patients the ability to digitally start and complete their Advance Directive documents.

Limitations - Our focus target segment is patients 65 years of age or older. Ned to build out a platform where they are comfortable with digital adoption.

Core Platform

The core platform for the Vynca Advance Care platform enabling users the ability to see a landing page Dashboard, Guided document workflows, e-signature workflows, and document review workflows.

AHA - American Hospital Association

American Hospital Association- Website link

Founded in 1898 the AHA is an advocacy group that focuses on 5,000 hospitals and 43,000 individual members. The AHA had a 100 year old hospital directory called the ‘AHA Guide’ that was distributed via mail to the hospital CEOs. In addition to the ‘AHA Guide’ they had a ‘AHA Hospital Statistics’ physical book that was provided via mail.

AHA Guide and AHA Hospital Statistics

Challenge - Build out an online version of the guide that would enable more users access vs a physical book has limited visibility to the CEO and other users in the hospital. Increase current commercial licensing from $800k annually. Increased raw data licensing fee and built ad-hoc tools to build internal and external teams their own reports that could be scheduled and delivered to internal/external stakeholders securely. AHA Guide link

Limitations - Hospital CEOs do not want another app to download or another password to remember. We want to allow all hospital members easy access to the new digital books.

AHA Data Query

Challenge - Built to enable internal, external and hospital CEOs the ability to create adhoc reports to measure their performance and or target hospitals for sales. This reporting tool was built to enable end users to leverage both internal proprietary AHA data, CMS and community census information. Once customized users can schedule for delivery to themselves and others.

Limitations - Create an intuitive and easy to use solution that would empower high utilizers with as much flexibility as possible. While also supporting low utilizers with a easy to use design to access company built reports and customize their own reports.

Google Sites

Report abuse