- Java
- Go
- Python
- JavaScript
- TypeScript
- Spring
- Express
- HAPI FHIR
- REST
- Server-Sent Events
- Vue 3
- HTML & CSS
- PostgreSQL
- MongoDB
DavidBelykh
Java full-stack engineer building dependable backends and the interfaces on top of them.
I work across the stack, from Java services and data pipelines to Vue and plain JavaScript frontends. Lately: healthcare interoperability with HL7 FHIR, load balancing in Go, and putting LLMs to work inside small, useful tools.
01
About
I like systems you can reason about: clear data flows, boring infrastructure, and code that the next person can pick up without a guided tour.
Most of my day is Java — services, APIs and the plumbing between them. On the side I use small projects to learn things properly: writing a load balancer from scratch in Go to understand health checks and request routing, wiring an LLM into a chess board so it can actually see the position, or building a HAPI FHIR server with an ingestion pipeline that turns messy clinical files into clean transaction bundles.
I care about the unglamorous parts: logging that helps at 3 a.m., Docker setups that start on the first try, and READMEs that tell you how to run the thing.
- Based in
- Lisbon, Portugal
- Currently
- Java Full-stack Engineer
- Languages
- Java, Go, Python, JavaScript
02
Selected work(05)
(01)
HAPI FHIR Server
An HL7 FHIR R4 server with an automated pipeline that ingests clinical files into PostgreSQL.
A HAPI FHIR server for HL7 FHIR R4, paired with a Python ingestion pipeline. The pipeline watches an input folder, normalises incoming clinical resources and posts them as FHIR transaction Bundles, so data lands in PostgreSQL through the server's own validation instead of around it.
The interesting part is the unhappy path: files that fail validation are moved to an error folder with logs that say why, processed files are archived, and the whole stack comes up with a single
docker compose up.- Continuous folder-watching ingestion with processed / error separation
- Resources sent as FHIR transaction Bundles for atomic writes
- One-command local stack with Docker Compose
Source 1(02)
Chess × LLM
A two-player chess board with an AI assistant that can actually see the position.
Started from Cloudflare's LLM chat template and grew into a split-screen app: a playable chess board on the left, a streaming AI chat on the right. Every message silently carries the current board as a matrix, so the model can comment on this position rather than chess in general.
The board does move validation, check and checkmate detection; answers stream back over Server-Sent Events from Cloudflare Workers AI.
- Board state injected into every prompt as a matrix
- Move validation with check / checkmate detection
- Streaming responses over Server-Sent Events
(03)
LoadBalancer
A load balancer written from scratch in Go to learn how request routing really works.
A from-scratch HTTP load balancer in Go, built to understand what sits in front of every production service: distributing requests across backends, checking which ones are healthy, and taking dead ones out of rotation.
- Round-robin routing across backends
- Health checks that drop unhealthy servers
Source 1
03
Experience
2024 — Now
Java Full-stack Engineer
Lisbon, Portugal
Building and maintaining Java services and the web frontends that sit on top of them, from API design to deployment.
Java / Spring / SQL / TypeScript
04
Capabilities
- Languages
- JavaGoPythonJavaScriptTypeScript
- Backend
- SpringExpressHAPI FHIRRESTServer-Sent Events
- Frontend
- Vue 3HTML & CSS
- Data & infra
- PostgreSQLMongoDBDockerCloudflare WorkersVercel