An Epic interface that works in test and fails in production is almost never a code problem. It is an identity, ordering or monitoring problem that the test data never exposed.
Charles Trinh architects Epic-facing interfaces from Las Vegas, Nevada: HL7 v2 feeds, FHIR endpoints, identity matching across enterprise and medical record numbers, and the monitoring that shows an interface is alive before a clinician notices it is not.

The first question is not which protocol, it is which identifier. An enterprise ID, a medical record number and a payer member ID are three different things, and a feed that assumes they are the same will merge two patients or split one.
The design names the source of truth for each identifier, the matching rule, and the behaviour when a match is ambiguous. In patient-facing work that rule is the difference between a record and a duplicate.
Most clinical traffic in the United States still moves as v2 messages. The work is to make those messages trustworthy: required segments present, identifiers mapped, order control codes handled, and acknowledgements that actually reach the sender.
A v2 interface with no negative acknowledgement path is an interface that fails silently, which is worse than one that fails loudly.
FHIR is where patient access, provider directories and newer application integrations land. It is a smaller surface than v2 and a stricter one: resource profiles, search parameters, pagination and authorisation all have to be right.
The usual pattern is v2 inside the enterprise and FHIR at the boundary, with a mapping document that states which resource field comes from which segment. That document is what a certification or audit review asks for.
An interface needs an owner and a signal. The signal is a message count, a queue depth and a latency number that someone watches; the owner is the person who responds when the number moves.
Without both, the first indication of failure is a clinician calling. With both, the failure is an alert that is closed before anyone in the building notices.
Test environments carry a small, tidy subset of patients with clean identifiers. Production carries duplicates, merged records, patients with two insurance plans and visits that span midnight.
The design has to be exercised against the messy case before go-live: a patient with two medical record numbers, a message that arrives out of order, an acknowledgement that never returns. A test plan that covers those cases costs a day and saves a month of clinical workarounds.
It starts with a scoping call of thirty to forty-five minutes. The point is not to sell; it is to establish what decision has to be made, what constrains it, and whether this is the right kind of work for one architect rather than a team. If it is not, the call ends with a referral.
If it is, the next artefact is a written plan: scope, sequence, price and the evidence each step will produce. Nothing starts before that plan is agreed in writing, and nothing in it is a surprise later. Plans change, and when they do the change is written down with its effect on price and date.
Delivery runs in weekly checkpoints. Each one ends with something that exists: a document, a design, a working system, a test result. That is deliberate, because a programme that reports progress without artefacts is a programme that cannot be audited, handed over or defended.
Two shapes. A fixed price for a defined deliverable - an assessment, a design, a due-diligence report - where the scope is clear enough to price honestly. Or a monthly seat for continuing advisory work, with a set number of days and a standing slot for decisions.
Invoices are monthly, net fifteen, and the terms are the same for every client. There is no percentage of savings, no success fee and no commission from a vendor, because an advisor paid by the seller is not an advisor.
Nothing formal. What helps is a one-page description of the problem in your own words, the names of the systems involved, the deadline that is driving the timing, and who else has to agree before anything changes.
If a vendor proposal is already on the table, send it. Reading someone else proposal is often the fastest way to find the question nobody asked, and it costs nothing to look.
Every engagement leaves a record: the map, the decisions, the trade-offs that were rejected and why. That record is what makes a change of staff survivable, what an auditor reads, and what a buyer inspects. It is also the part most consultancies keep for themselves.
The record settles arguments. When a decision is questioned a year later, the note that states what was known, what was chosen and what it would cost to change is usually the only thing that ends the discussion.
Las Vegas health systems and the vendors that serve them run Epic alongside a long tail of ancillary systems. Local presence matters when an interface problem needs a room, a whiteboard and the analyst who owns the feed.
Scope is a fixed number of hours per month or a defined project, agreed in writing before work starts. Every engagement carries a written plan, a weekly checkpoint, and a closing pack that states what was built, what was verified and what is left.
Charles Trinh is a Las Vegas-based technology architect and fractional CTO/CIO serving health systems, payers, enterprises and government programmes in Nevada and nationally. Every discipline below is offered in ten engagement forms - consultant, architect, advisor, contractor, specialist, expert, services, support, solutions and consulting - so an engagement can be a review, a build, a retainer or an interim role.
The full catalogue, every discipline in all ten engagement forms, sits on the Las Vegas, Nevada services index. Delivery is remote-first from Las Vegas with on-site coverage across Clark County, and the same ten disciplines carry to engagements anywhere in the United States and in 264 countries.
Serving customers in Columbus, Ohio, United States and worldwide.
Teams searching for Epic interface architecture near Las Vegas usually need someone who has read a specification, traced an identifier and sat with the analyst who owns the feed.
No. The architecture, mapping and monitoring design can be produced from specifications, test messages and interviews with the analysts who own the interfaces.
Yes. That is the most common shape: a vendor needs a defensible interface design and the health system needs to review it.
Yes, field by field, with the source of truth named for each.
Yes, including API design, consent and the evidence a review expects.
Two to three weeks for a defined set of feeds.
trinh.io · Services · Las Vegas, Nevada · charles@trinh.io · 702-559-3489