Distributed systems · NATS/JetStream event pipelines · LLM & agent infrastructure (RAG / GraphRAG)
Hi, I'm Matvey — a Go backend engineer who likes hard distributed-systems problems and the messy edges of putting LLMs into production. Most of my work lives where event-driven backends meet AI: building pipelines that orchestrate language models reliably, and modeling infrastructure as graphs so software can reason over it.
🤖 LLM & agent systems. My MSc thesis built a RAG + GraphRAG agent over an infrastructure graph (grade 9/10), demonstrated at HighLoad++ 2024 — abstract here. I work on tool-calling pipelines with self-correcting validation loops, structured-output control, and getting clean, reliable data out of models that don't always cooperate.
⚙️ Event-driven backends. Multi-stage NATS/JetStream pipelines — fan-out, worker pools, back-pressure, exactly-once processing under at-least-once delivery, and watchdog/reconciliation for work that gets stuck.
🕸 Graph / CMDB modeling. Currently at Foliage, building the backend of a distributed infrastructure-discovery platform: a heterogeneous graph fed by ~19 discovery connectors (SNMP, Pacemaker, Kubernetes, virtualization, S3), exposed over GraphQL/REST, with parallel reconciliation and adaptive rebuild scheduling.
🧠 Also: MSc student in Systems Analysis at HSE University; into concurrency patterns, message brokers, and MCP servers.
Backend & APIs: Go · gRPC · Protocol Buffers · REST · WebSocket · concurrency (worker pools, pipelines, context propagation)
Messaging: NATS JetStream · event-driven pipelines · statefun/actor model
Data: PostgreSQL (pgx, raw SQL) · Redis · graph data modeling
LLM / AI: RAG · GraphRAG · AI agent systems · LLM tool-calling pipelines · structured-output control · MCP servers