Sergiy Yevtushenko

Founder, Pragmatica Labs | Application Operating System for Java | 35 years production experience

After 35 years building production systems -- from embedded firmware to trading platforms -- I founded Pragmatica Labs to solve the problems I kept seeing: fragile distributed architectures, unpredictable codebases, and operational complexity that grows faster than the applications it supports.

The result: Aether -- a fused application runtime that handles everything between your business logic and production. One binary, one config file, one system to operate. And JBCT -- a coding methodology that turns Java code into executable business process specifications.

Pragmatica Labs

Website: pragmaticalabs.io

Aether

Fused Application Runtime for Java | Repository | Demo

Write business logic as slices -- Java interfaces with methods. Deploy into a runtime that handles orchestration, scaling, resilience, streaming, security, and observability. No framework, no external infrastructure dependencies.

What's built in:

• Rabia consensus + two-layer topology (core + SWIM worker groups) -- proven at 12 nodes, zero failures
• QUIC transport for all inter-node communication -- first Java distributed runtime on QUIC
• HTTP/3 with dual-stack H1+H3 for application and management APIs
• In-memory streaming -- ordered, replayable, consumer-paced. No external message broker
• Cloud providers (AWS, GCP, Azure, Hetzner) -- raw REST APIs, no vendor SDKs
• Deployment strategies -- canary, blue-green, A/B testing with auto-rollback
• Schema migrations -- Flyway-style, consensus-coordinated, gated on readiness
• JWT/JWKS auth, RBAC (ADMIN/OPERATOR/VIEWER), operational audit trail
• Async PostgreSQL driver with built-in pipelining -- no external connection pooler
• Declarative cluster management -- bootstrap, scale, upgrade from a single TOML file
• Predictive autoscaling -- ML-based (ONNX) with 11-metric feature vector
• 500+ tests including 4-hour soak test with chaos injection

Performance: 8K req/s at sub-5ms p95 with real PostgreSQL on a laptop. Scales to 15K req/s.

Business Source License 1.1 → Apache 2.0 after 4 years.

JBCT

Java Backend Coding Technology | Repository | Documentation | Book

A methodology and tooling for writing deterministic, AI-friendly Java code. Six structural patterns that map directly to BPMN constructs -- making code an executable business process specification.

• Four return types -- T, Option<T>, Result<T>, Promise<T> cover every method signature
• Six patterns -- Leaf, Sequencer, Fork-Join, Condition, Iteration, Aspects. Each maps to a BPMN construct.
• BPMN bridge -- code structure mirrors business processes. Readable by non-technical stakeholders.
• CLI + Maven plugin -- automated validation with zero false negatives
• 3.5% complexity ratio -- measured by scc, compared to typical 12-18% in Spring Boot applications
• Parse, don't validate -- invalid states are unrepresentable

Monorepo

pragmaticalabs/pragmatica -- 230K lines of Java, 1,732 files

Module	Purpose
core/	Result, Option, Promise -- functional primitives with zero dependencies
integrations/	Jackson, Micrometer, PostgreSQL, SMTP, cloud providers (AWS/GCP/Azure/Hetzner), XML, HLC
jbct/	CLI, Maven plugin, slice annotation processor
aether/	Distributed runtime, Forge simulator, Ember embeddable mode

Articles

• We Should Write Java Code Differently: Let's Get Practical (2026)
• Nanoservices: Alternative to Monoliths and Microservices (2019)
• We Should Write Java Code Differently (2021)
• Introduction to Pragmatic Functional Java (2019)

Get In Touch

• Website: pragmaticalabs.io
• LinkedIn: sergiy-yevtushenko