10/03/2026 updated


100 % available
Senior Java Engineer | Freelancer & Contractor | AI-powered Backend Systems
Cluj-Napoca, Romania
Worldwide
BS in Computer ScienceAbout me
With over 8 years of experience working in international teams, I specialise in AI-powered backend systems and the controls that make them safe to run in production: agent orchestration, sandboxed execution, policy enforcement and streaming data platforms.
AI AgentsArgo CDRAGClean ArchitectureLLMHelmBedrockSAFeChatGPTJava (Programming Language)Spring SecurityAgile MethodologyAmazon Web ServicesBig DataUnix
Java Backend Development
Expertise in Java (versions 11 through 25), Spring Boot, Spring Data, Hibernate, and JavaEE, with extensive experience building scalable, multi-tenant backend systems and microservices for international clients across various industries.
AI-powered Systems & Agent Orchestration
Hands-on experience integrating AWS Bedrock (Claude), Codex, and Claude Code to build AI-driven evaluation systems, agent orchestration pipelines, and sandboxed code execution environments with enforced policy controls.
Cloud & Container Infrastructure
Proficiency in Docker, Kubernetes, AWS, GCP, IBM Cloud, and OpenShift, including designing and deploying containerized pipelines, managing CI/CD workflows with Gitlab CI/CD, Helm, and ArgoCD, and migrating large-scale workloads to cloud environments.
Streaming Data Platforms
Experience with Kafka, Kafka Streams, and Apache Flink for building event-driven, real-time streaming architectures, including watermark-based event-time processing and broadcast state management.
Database Management
Strong knowledge of PostgreSQL, MySQL, MongoDB, Redis, and Oracle Database, including advanced techniques such as pgvector semantic routing, SKIP LOCKED job queues, and bulk soft-deletion optimizations.
Messaging & Integration
Practical experience with RabbitMQ, Node-red, and Control-M for asynchronous messaging, job scheduling, and system integration across distributed environments.
Testing & Quality Assurance
Skilled in JUnit 4 and 5, Arquillian, Spock, and Testcontainers for writing comprehensive unit, integration, and contract tests to ensure software reliability and correctness.
Build & Dependency Management
Proficient in Maven and Gradle for project build automation, dependency management, and multi-module project structuring in enterprise Java environments.
Python & Big Data
Working knowledge of Python, PySpark, and Hive for data processing, scripting, and big data analytics tasks within cloud and on-premise environments.
Security & Policy Enforcement
Experience designing secure pipelines with credential scoping per phase, secret scanning, path denylists, git-config tamper detection, and fingerprinting to enforce zero-trust execution environments for AI agents.
Experience
Professional experience5 - 10 years
Freelancer since2021
Languages
EnglishFluentRomanianNative speaker
Project history
Working at Tepsivo, a scale-up in the pharmacovigilance industry. Architected and implemented a dynamic, multi-tenant state machine enabling runtime extension of workflows without redeployment. Optimized bulk soft-deletion by refactoring legacy logic and moving deletion handling to the database layer, reducing deletion time for 1M records to approximately 11 seconds. Integrated AWS SES with AWS Bedrock (Claude) to generate contextual email content based on structured parameters and ingested documents. Built an AI-driven evaluation system using AWS Bedrock Knowledge Bases, aggregating and querying data from multiple scraped articles to assess expectancy of domain-specific inputs.
Self-hosted service that turns a Slack or Discord request into a reviewed merge request, running AI coding agents in isolated containers behind enforced policy controls. Designed a four-phase container pipeline scoping credentials per phase so the AI agent executes with zero Git or Slack tokens present. Built a durable PostgreSQL job queue with SKIP LOCKED atomic claims, leases, heartbeats, idempotency keys and automatic recovery of orphaned jobs. Added pgvector semantic routing to select the target repository from a natural-language request. Made the coding agent provider-neutral with Codex or Claude Code behind one configuration switch.
Redesigned a fragmented multi-service pipeline into a Kafka-based streaming architecture, consolidating the number of services involved and removing coordination overhead. Implemented event-time processing with watermarks and 10-second tumbling windows so decisions reflected when measurements actually occurred. Closed gaps in incoming data by joining against a broadcast stream of last-known values and distributed configuration and decision rules via Flink broadcast state. Filtered invalid quality codes and enriched events before publishing downstream, delivering decisions to end users in real time.