NutriSense
- Go
- Flutter
- LLM
A shipped mobile product: Go backend, Flutter client, and an assistant with hard limits on what it is allowed to claim.
Viktor MarymorychBackend & Systems Engineer
I build production systems that hold under failure and load.
Rust, Go and Python backends. Real-time and event-driven architecture. AI agents and MCP integrations.
Selected work
Products in production, systems delivered for partners, and the internal tooling my team runs on. Each one is here for the engineering problem it had to solve.
RustSTT
- Rust
- WebSocket
- CUDA
- Metal
Real-time speech-to-text in Rust: streaming over WebSocket, several recognition architectures, one service that runs on CPU, CUDA or Metal.
Vestry
- Rust
- MTProto
- gRPC
- Protocol Buffers
A distributed Telegram-facing system: four Rust backend services talking over gRPC, with agent-adjacent processing behind the protocol boundary.
More work
Cortex
Pre-release. Source is private.
Network reconnaissance and systems tooling, still ahead of release.
Aelyx
Internal engineering tool, not a generally available product. Source is private.
Engineering intelligence for a codebase: Git-aware context, durable project memory and architecture knowledge, exposed to AI tools over MCP.
Rust / MCP / Git / Developer tooling
Lexora
Live product. Source is private.
A production Rust backend with an event-driven core: transactional outbox, background workers and a React front end, running on Cloud Run.
Rust / React / Cloud Run / Background workers
Sentinel Solar
Deployed at installations. Source is private.
Edge telemetry for solar installations: an always-on local backend next to the inverter, not a cloud-only service.
Go / PostgreSQL / Docker / Edge
Internal tooling
Not generally available software. Source is private.
Web Aegis and Blockchain Aegis: proprietary tooling the Red Sentra team uses for its own security-related delivery work.
Web Aegis / Blockchain Aegis
What I build
Four kinds of problem I take on, and what the work actually involves.
Real-time and event-driven systems
WebSocket transport, Redis event processing, transactional outbox, background workers. Retries, idempotency and back-pressure designed in from the start, so partial failure stays partial.
AI agents and LLM systems
MCP servers and tool integrations, guarded agent behaviour, context and memory design. A model is a component with a failure rate, so there is always a deterministic path when it is wrong or unavailable.
Distributed backends
Rust and Go services talking over gRPC and Protocol Buffers. Service boundaries, tenant isolation, schema and migration strategy, and the operational story for running it after launch.
Edge and performance work
Streaming speech-to-text in Rust across CPU, CUDA and Metal. Always-on local backends that sit next to the hardware they talk to and keep working when the network does not.
Where I work
Two current roles. Both hands-on.
Red Sentra (opens in a new tab)
December 2025 to present Remote
Founder & Systems Engineer
Red Sentra is a remote engineering team of eight, including me. We build and operate our own products, deliver commissioned systems for external partners, support what we ship once it is live, and maintain internal engineering and security tooling. I stay technical: architecture, Rust and Go services, and the parts of delivery nobody else wants.
- Own products in production, plus commissioned delivery for external partners
- Long-term production support for systems the team has shipped
- Internal engineering and security tooling built and maintained in-house
Efficlose
December 2025 to present Remote
Backend Engineer
Product engineering around an AI assistant that joins meetings and takes the administrative follow-up off people's hands. I work on the real-time path and the integrations that make the output usable inside a company's existing tools.
- Real-time capture path: Rust speech-to-text integration, WebSocket transport, Redis event processing
- LLM and CRM integrations, including MCP tooling
- Migrations, infrastructure, security and production reliability
Before that
Contract backend engineering and offensive-security work, most of it under NDA. Bachelor's degree in Cybersecurity and System Engineering, Odesa Polytechnic, 2025.
How I approach a system
Five positions I hold. All of them show up in the code.
Design for the failure case first
Timeouts, retries, idempotency and back-pressure belong in the first design, not in a later patch. A service that only works on the happy path is not finished.
Security is an architecture decision
My offensive-security background shows up in where trust boundaries sit, how tenants are isolated, how services authenticate to each other, and how secrets, encryption and audit trails are handled. It is a property of the design, not a phase at the end.
Make behaviour observable
If I cannot reconstruct what a system did from its logs, metrics and events, I do not consider it operable. Debugging in production is a design requirement.
Keep models on a leash
Agents get explicit tool boundaries, validated inputs and outputs, and a fallback that does not need the model to be right. Non-determinism is acceptable in the answer, not in the plumbing.
Ship it, then support it
Most of what I have built is still running and still mine to fix. Owning the on-call side changes the decisions you make on day one.
Have a system that has to hold up?
Backend and distributed systems, AI agent integrations, or a production system that is not behaving. Write directly and describe the problem.
