5+ years managing enterprise infrastructure, cloud systems, and mission-critical incidents at scale.

HL7/FHIR Healthcare Data Pipeline

2026 · Data Engineering · Python · SQL · FHIR

Problem: Healthcare data still runs largely on HL7 v2, a decades-old pipe-delimited messaging standard, while the industry is migrating toward FHIR (JSON-based). Vendor feeds don't always send clean data — malformed dates, missing required fields, garbled segments — so anything touching clinical data has to validate defensively before it's trusted downstream.

Insight: Not all failures are the same. A message that isn't valid HL7 at all, a message that parses fine but is missing clinical data, and a message that fails at the database layer are three different problems with three different fixes. Treating them as one bucket makes root-causing a vendor issue much slower than it needs to be.

Solution: Built an end-to-end pipeline that generates synthetic ADT (admit) and ORU (lab/vitals) HL7 v2 messages — some intentionally malformed — parses and validates them, loads clean records into a relational database, and transforms valid messages into FHIR R4-style Patient/Encounter/Observation JSON bundles. Anything that fails is dead-lettered with its raw message, the stage it failed at, and the specific reason — nothing is silently dropped.

Technical highlights:

• HL7 v2 parsing and validation layered separately, so parse failures and data-quality failures are tracked and reported independently
• SQLAlchemy models (Patient, Encounter, Observation, FailedMessage) with idempotent upserts keyed on MRN and visit number
• Backend-agnostic by design — the same code runs against SQLite, PostgreSQL, or SQL Server via a single connection string
• FHIR R4 transformation layer producing Patient, Encounter, and Observation resources as JSON bundles
• 14 pytest tests covering parser edge cases, every validation rule, and full pipeline runs
• CLI with configurable message volume, malformed-message rate, and random seed for reproducible runs

Results:

• A 200-message run resolves roughly 175 messages successfully and dead-letters ~25 with categorized, human-readable reasons
• Every stage — generate, parse, validate, load, transform — is independently unit-tested

What this demonstrates: End-to-end data engineering across ingestion, validation, storage, and transformation; fluency in both the legacy HL7 standard and the FHIR standard replacing it; and the defensive, audit-friendly engineering that regulated healthcare data requires.

PulseList — Task Management

2026 · Full-Stack Web Development · JavaScript · Firebase

Problem: Most task apps treat deadlines as static dates. You see a list of due dates, but you don't immediately know what needs attention in the next hour vs. next week.

Insight: Urgency is about time remaining, not just the date itself. Real-time countdown timers — not just due dates — tell you what to focus on next.

Solution: PulseList puts countdown timers front and center. Every task displays live time remaining, with color signals that shift from green → yellow → red as deadlines approach. Supports multiple views: a focused Today dashboard, a full task grid, and a timeline for longer-term planning.

Technical highlights:

• Cloud sync via Firestore for cross-device access
• Real-time countdown updates without page refresh
• Firebase Auth + Google Sign-In for optional authentication
• Fully responsive with mobile sidebar navigation
• Google Calendar integration for exporting tasks
• Semantic HTML and ARIA labels for accessibility

What this demonstrates: Full-stack web development, cloud-synced architecture, product thinking, and production-ready code with accessibility and error handling built in.

Anchor — Personal Finance Simplified

2026 · Full-Stack Web Development · JavaScript · Firebase

Problem: Financial apps are either too complex, too simplified, require account linking and data sharing. None of them start with what most people actually need to know.

Insight: Most people only need a few core metrics: How much am I spending? Am I on track with my budget? What's my net worth? When are my bills due? How close am I to my savings goals?

Solution: Anchor is a lightweight finance app that syncs to Firebase for cross-device access, and covers the essentials: accounts, transactions, budgets, goals, and bills — with net worth calculated as assets minus liabilities.

Technical highlights:

• Cloud sync via Firestore for cross-device access
• Financial calculations: net worth, budget tracking, interest accrual
• Data modeling: relationships between accounts, transactions, budgets, and goals
• Firebase Auth for optional cloud sync
• Responsive across mobile, tablet, and desktop

What this demonstrates: Complex business logic, financial data modeling, full-stack development, and user-centric design.

Patient Intake Form Development — Nextech Systems

2025–Present · Healthcare IT · Forms Development · Microsoft Office Developer Tools

Problem: Manual patient intake is one of the biggest sources of friction in a medical practice. When patients fill out unstructured documents, front desk staff re-enter that data by hand — introducing errors, slowing check-in, and creating compliance risks in a HIPAA-regulated environment.

Insight: A well-structured digital form isn't just a convenience — it's a data pipeline. If the fields are defined correctly from the start, the data flows cleanly into the system without manual intervention. The form is the first step of the clinical workflow.

Solution: Designed and developed structured patient intake forms at a volume of 5–10 forms per day using Microsoft Word and Excel with the Developer toolkit — building forms with defined data fields, content controls, dropdowns, checkboxes, and input validation to standardize how patient information is captured across the practice. Forms were built to map directly to the data fields used in Nextech's EHR platform.

Technical highlights:

• Built and maintained 5–10 structured intake forms daily using Word and Excel Developer mode
• Defined and mapped data fields to align with downstream Nextech system requirements
• Applied input validation and field constraints to reduce capture errors at point of entry
• Designed with HIPAA compliance in mind — controlling what data is collected and how it flows
• Iterated on forms as clinical workflows and product requirements evolved

Impact:

• Measurably reduced data entry errors across patient intake workflows
• Saved front desk staff significant manual re-entry time per patient
• Standardized intake data quality across the practice, improving downstream system accuracy

What this demonstrates: Data field design and mapping in a clinical environment, HIPAA-aware development, healthcare IT domain knowledge, and high-volume production output with real operational impact.