Volodymyr Galushka
AI, backend, and DevOps engineer with 10+ years shipping production systems — Python and Django services, RAG and LLM pipelines, and the GCP infrastructure underneath.
Summary
I make systems talk to each other: enterprise chatbots, Django backends, and the infrastructure they run on.
Most recently at KBC, a Belgian bank working in a SAFe setup, I delivered 75+ business cases as RASA conversational-AI workflows — dialogue and intent design, Python action servers, and integrations with banking backends, including debugging across Python and Java services.
Before that I did Python/Django backend work: a partner-facing portal for high-voltage grid companies, WhatsApp messaging over RabbitMQ, Elasticsearch full-text search, and data layers from PostgreSQL and Redis to Neo4j. On the infrastructure side my production experience is GCP — GKE, Docker, Terraform, 20+ Jenkins CI/CD pipelines, and ELK/Loki monitoring with the runbooks behind them. AWS and Azure were brief operating stints, not deep production work.
On the AI tooling side I work with LangChain and the OpenAI API, and I've chosen vector databases the unglamorous way: a hands-on evaluation moved us from Weaviate to Qdrant. I'm now going deeper into production LLM engineering — a RAG system with real evals, observability, and documented tradeoffs.
What I do best
- Design and integrate conversational AI: RASA dialogue systems, intent training, and action servers wired into enterprise systems
- Build Django and FastAPI backends with clean SQL, NoSQL, graph, and vector data layers
- Operate what runs on GCP — GKE, Terraform, CI/CD, monitoring — and debug across the app/infra boundary
- Choose tools on evidence: hands-on evaluation, honest tradeoffs, written findings
I work with teams across Europe. I'm currently open to Applied AI, backend, and cloud/DevOps roles.
Experience
Software Engineer
- Delivered 75+ business cases as RASA chatbot workflows and intents for a Belgian bank
- Built generalized Python action-server classes adopted across teams
- Resolved 100+ production bugs across Python and Java services, including debugging across service boundaries
- Worked in a SAFe setup with Scrum Masters: sprint planning, backlog refinement, and PI ceremonies
Backend Engineer — Python/Django & LLM Pipelines
- Built a periodic mailbox-processing pipeline (Django, Celery, PostgreSQL, Redis) with idempotent re-parsing, per-field timestamps, and merge rules so newer data wins
- Hardened LLM extraction with LangChain and the OpenAI API: validation and cleanup rules, temperature 0, and lowered top-k/top-p for deterministic output
- Added structure-validated summarization with automatic fallback to a local extractive summarizer, so triage survives model or API failures
- Selected the vector store through hands-on evaluation, migrating production from Weaviate to Qdrant
- Built per-user observability: Loki logs tagged by email, Grafana-filtered queries, and staff pages to replay parser runs and prompt tests
- Built the document register (versioned documents, release metadata, audit trail) and rollback-safe uploads keyed by UUID in Azure Blob storage
- Implemented the multi-company permission layer and planning module; mentored a second developer across three Docker Compose environments (prod, debug, test)
Cloud & DevOps Engineer
- Operated production infrastructure on GCP — Compute Engine, GKE, networking, and storage for client workloads
- Cut release cycle time by 30% by building 20+ Jenkins CI/CD pipelines for build, test, and deployment
- Reduced environment bootstrap time by 30% by codifying infrastructure in Terraform
- Configured and maintained a production GKE cluster — node pools, workloads, and zero-downtime upgrades
- Containerized 20+ Node.js applications, eliminating environment drift between stages
- Deployed ELK Stack logging and monitoring, raising issue-resolution speed by 25%
- Automated SQL and NoSQL backup pipelines to on-premises hosts and wrote the runbooks, cutting support tickets by 20%
Software Developer
- Built the Qt/C++ client for dome control, including synchronized multi-device playback and a standalone content-production and encryption application
- Designed a custom content-encryption scheme with libsodium to protect licensed projection content at customer sites
- Wrote server daemons for Gentoo Linux in Python and C++ that kept long-running show services responsive
- Programmed embedded controllers (ESP8266/ESP32/Arduino/STM32), including multi-device LED lighting synchronization
- Upgraded the iPad control app in Swift, reducing UI crashes by 30%
Embedded Developer
- Ported and developed router firmware (Tomato/Sabai on Cavium Octeon hardware)
- Implemented the HDLC protocol for mobile base-station antenna tilt control, improving command reliability
- Built early device-cloud connectivity components for Cirrent, with initial device communication delivered in two weeks
Software Developer
- Developed ground-station segment software in C# and WPF
- Implemented basic AX.25 protocol procedures for the nanosatellite microcontroller in pure C
- Researched packet-error-rate reduction based on variable packet length
Education
Master's Degree, Computer Software Engineering
National Technical University of Ukraine "Kyiv Polytechnic Institute" · GPA 4.6 / 5.0
Bachelor's Degree, Computer Software Engineering
National Technical University of Ukraine "Kyiv Polytechnic Institute" · GPA 4.6 / 5.0