Denis Sheremetov
Denis Sheremetov is the CTO of Onix Systems. He leads engineering across the company's healthcare work: interoperability, clinical systems, remote patient monitoring and medical device software built under an IEC 62304 process.

What Denis works on
He writes on the technical core of healthcare engineering: FHIR and HL7v2 integration, clinical decision support that clinicians actually adopt, and shipping compliant software without slowing delivery to a crawl.
- Healthcare interoperability (HL7 FHIR, HL7v2)
- Clinical decision support
- Software as a Medical Device (SaMD)
- Healthcare AI and ML
Articles by Denis
- Epic integration that supports read and limited write: how write-back works
Reading from Epic is predictable. Writing back is a set of specific Create and Update APIs, each with its own rules, and every health system decides which of them your app may use. - Patient portal software that integrates with Epic: how the connection works
A patient portal connects to Epic as a patient-facing app: patients sign in with MyChart, authorize the data your portal requests, and the portal reads and writes through the APIs registered to its client ID. - athenahealth API integration: what the work involves, step by step
An athenahealth API integration connects your product to a practice's athenahealth environment through standard FHIR APIs, athenahealth's own platform APIs, or both, with the practice's approval. Here is how to scope it. - The easiest way to connect to major EHRs: direct APIs, aggregators or national networks
There are three ways to connect a product to major EHRs: build directly on each vendor's APIs, buy one API from an integration platform, or retrieve records through a national network. Each is easiest for a different job. - How to sync lab results and notes into the EHR without losing data
Syncing lab results and notes into an EHR means matching each item to the right patient and order, mapping codes and units, handling corrections as updates, and filing notes where clinicians will read them. - Self-hosted Mirth Connect vs an integration platform: run your own engine or buy
A self-hosted integration engine such as Mirth Connect gives you full control of every interface and makes you responsible for running it. An integration platform sells you maintained connections. Since Mirth's 2025 licensing change, the engine option has its own new choices. - SMART on FHIR app launch explained: launch types, scopes and tokens
SMART on FHIR app launch is the OAuth 2.0 flow an app uses to get a token for an EHR's FHIR API, either launched from inside the EHR or started on its own. The steps, the scopes, and the rules that trip teams up. - Oracle Health (Cerner) FHIR API: A Developer's Guide to Building on Millennium
How to build on the Oracle Health Millennium FHIR R4 APIs: code Console registration, the open and secure sandboxes, SMART authorization details, system accounts, bulk data and the service requests a health system files to turn your app on. - Bulk FHIR Export: How to Pull Population Data From an EHR
Bulk FHIR ($export) is the standard way to pull data for a whole population from an EHR or payer. Here is how the asynchronous flow works, which parameters matter, how backend authorization works and how to build a pipeline that survives production. - Mobile Check-In Software and EMR Integration: How the Intake Flow Works
Mobile check-in only saves front-desk time if what patients enter lands in the EMR. Here is the intake flow step by step, the FHIR resources and events behind it, insurance verification, and the integration problems to plan for. - Hybrid Patient Portal Solutions: Web, a Mobile App, or Both?
A hybrid patient portal serves the same accounts, data and permissions through a web app and mobile apps. The work is deciding which tasks go to each channel, meeting app store, NIST and accessibility rules on both, and knowing when web alone is enough. - EMR-Agnostic Patient Portal: One Portal Across Many EHRs
An EMR-agnostic patient portal keeps its own accounts and design and reaches each EHR through an adapter. What certified EHR APIs guarantee, what they leave out, and how to structure the portal so that adding an EHR means adding an adapter. - HIPAA Secure Patient Messaging: Building Messaging Into Your Health App
Secure messaging inside a health app means patients and care teams exchange messages behind sign-in, with every message treated as ePHI. How the HIPAA Security Rule maps to messaging design, how to route and record messages, and when portal messages are billable. - House Call Scheduling Software: How to Plan and Route Provider Visits
House call scheduling software turns a list of patients due for a visit into a route for each provider, under real constraints: visit frequency, a daily cap, time windows and travel. Planning and routing work best as separate steps, patient data should stay out of the mapping service, and each visit record needs specific fields for billing. - HIPAA-Compliant Insurance CRM: How to Handle PHI in a Sales and Service System
An insurance CRM becomes a HIPAA system the moment it holds a member's plan, doctors or medications. From there, your role as covered entity or business associate decides who signs which BAA, the Security Rule sets the safeguards, and Salesforce and HubSpot each limit where PHI can go. - SaMD vs SiMD: How to Tell Which One Your Software Is
Software as a medical device performs a medical purpose on its own. Software in a medical device is part of the hardware and needed for it to work. FDA's examples, the gray zone between the two, the software that is not a device at all, and what changes for your team either way. - 510(k) Software Documentation: What to Prepare Before You Submit
A 510(k) for a device with software needs the standard premarket notification content plus a software package sized to the device's risk. The package has ten elements, Basic and Enhanced documentation differ in three of them, and the work goes most smoothly when you produce it while you build. - ISO 14971 Software Risk Management: Building the Risk File in Practice
ISO 14971 sets the risk management process for medical devices, software included. How a software team builds the plan, hazard analysis, risk controls, residual risk evaluation and report that FDA expects, and why the file stays open after release. - PCCP for AI Medical Devices: How to Plan Model Updates Before Clearance
A predetermined change control plan lets an AI-enabled device change after authorization without a new submission for each change, if FDA reviewed the plan up front. What goes in its three sections, which changes fit, what happens after authorization and how to build the pipeline for it. - HIPAA Compliant Hosting on AWS: Eligible Services and Who Secures What
AWS can host protected health information, but no AWS account is HIPAA compliant on its own. You accept the AWS BAA, keep PHI in the eligible services, and own the safeguards on your side of the shared responsibility line. - Clinical AI Assistant With Citations: Building Answers Clinicians Can Check
A clinical AI assistant with citations ties every statement to the exact passage in the chart or guideline it came from, so a clinician can check it in one click. Here is how Claude's citations feature works, how to feed it patient records and what citations alone do not solve. - Running LLMs Under a HIPAA BAA: Amazon Bedrock, Anthropic and What Each Covers
Before PHI goes into a prompt, someone has to sign a business associate agreement and you have to stay inside what it covers. Here is how the AWS BAA applies to Amazon Bedrock, what Anthropic's own BAA covers feature by feature, and the retention, logging and region settings that decide where PHI ends up. - Medication Decision Support Software: Putting Medication Checks Into the Ordering Workflow
Medication decision support checks an order against the patient's other drugs, allergies, conditions and labs before it is signed. Here is what the checks cover, what certified EHRs already do, where your checks should fire and when FDA treats them as a device. - Clinical Decision Support in the Clinical Workflow: Where CDS Fits
Clinical decision support works when it shows up at the step where a decision is made, in a form that suits that step. Here is a map of the clinical workflow, the kind of CDS that fits each stage, the integration options and how timing changes the regulatory picture. - How to Evaluate LLM Accuracy in Healthcare Before Release
Evaluating a healthcare LLM before release means defining what accurate means for your task, building a test set from real questions and hard cases, grading answers with code, a second model and clinicians, and setting a release gate you re-run on every change. The method below follows guidance from Anthropic, AWS, NIST and FDA. - Redox vs Health Gorilla vs 1upHealth: choosing a healthcare integration platform
Redox offers managed integration with your customers' EHRs, Health Gorilla a national clinical network, and 1upHealth FHIR-native data for health plans. Picking the wrong one is slow and expensive to undo, so start from the data flow you need. - FHIR vs HL7v2: which standard for which integration
HL7 v2 carries the event feeds inside hospitals; FHIR is the API that certified EHRs and health plans must now expose. Each kind of integration fits one of them better, and most US projects will run both for years. - Remote Patient Monitoring Without Alert Fatigue: Alert Design and the Billing Rules Behind It
Alert fatigue sets in when an RPM program sends care teams more alerts than they can act on. How to tier and route alerts, how the Medicare RPM codes and rules should shape the design, and which metrics show fatigue early. - What a HIPAA audit actually looks like: the process, the documents and a checklist for software teams
An OCR HIPAA audit starts as a document request with a short deadline. How the audit program works, which Security Rule controls it has tested, and the evidence a product or engineering team should be able to hand over. - Patient Portal Adoption: What the National Data Shows and What to Measure
Most US patients are offered a portal and most of them log in at least once a year. The harder part is getting them to come back. Here is what federal survey data says, the metrics worth tracking, and when a white-label patient portal makes sense.