Yurii Serhiichuk

Own Travelshift's infrastructure end-to-end — platform, reliability, security, developer experience, and AI-assisted engineering.

• Cut multi-cloud spend ~65% — through spot instances, right-sizing over-provisioned databases, Redis, and Kubernetes workloads, tuned autoscaling policies, leaner setups, cost-center tagging for visibility, and consolidating databases into a single cloud.
• Led a four-years-stalled AWS→GCP (GKE Autopilot) migration, delivered in under 8 months with zero customer-visible downtime across a gradual 1%→100% traffic shift and bidirectional database cutover.
• Operate a production AWS + GCP platform end-to-end — on AWS, EC2 Auto Scaling groups, ECS/Fargate, Lambda, RDS, DynamoDB, ElastiCache, and VPC/Route 53/IAM networking and access; on GCP, GKE Autopilot with Istio and KEDA — cloud infrastructure as code in Pulumi (migrated from Terraform) so any engineer can change it safely.
• Rebuilt incident management and on-call on PagerDuty — from 2 high-priority services to full per-service, database, and cache coverage with automated alerting, on-call escalation routing, and post-mortem workflows.
• Automated web protection and DDoS mitigation at scale with Signal Sciences (Fastly next-gen WAF) and the Fastly edge, VCL rate-limiting, and a JA3/JA4 TLS-fingerprinting feedback loop that auto-enforces security rules.
• Replaced long-lived cloud service-account keys with OIDC federation — GitHub Actions and Bitbucket as identity providers via Workload Identity Federation — for short-lived, keyless CI/CD credentials.
• Own the secure SDLC — automated security scanning (gitleaks, AI-assisted review) in CI and centralized secrets management across Doppler, Google Secret Manager, and AWS Secrets Manager.
• Run CI/CD across GitHub Actions, CircleCI, and Google Cloud Build.
• Deploy Kubernetes workloads three ways — Pulumi, kubectl with Kustomize, and home-grown Helm charts for standardized infrastructure.
• Wrote Go across the platform — a Kubernetes operator that stitches Istio VirtualServices, internal stitching/edge proxies, and open-source tooling (Open Knowledge Format).
• Standardized observability and SLO/SLI tracking on OpenTelemetry — collectors with tail sampling across GCP and AWS — and consolidated the tracing backend from Grafana Cloud to SigNoz for cost and usability.
• Drove org-wide adoption of AI-assisted engineering — standardized on Claude Code, rolled out automated AI code review, and built MCP integrations for internal systems.
• Lead the DevOps team and own cross-team architecture reviews and solution-design sign-off for new initiatives — driving architecture decisions and code review across engineering teams.

• Built the company's first AI assistant — an interactive AI travel guide combining pre-processed itinerary data with LLM APIs.
• Replaced the Google Maps routing API with a self-hosted OSRM service — car, foot, and a rail profile our team built and curated, across Iceland and Europe — later adopted company-wide for significant cost savings.
• Architected the AI/data-generation platform for tight cost control — every subsystem scales to zero or runs fully serverless (Cloud Run, Cloud Functions, GCP Workflows) on reusable Pulumi (Python) components.

Select client engagements alongside full-time roles — embedded builder-consultant work on application development, AI systems, and cloud migration.

• AgencyEZ (US insurance platform) — ongoing architecture lead on AWS (Elastic Beanstalk, EC2 Auto Scaling, RDS/Aurora, SQS/SNS, ElastiCache, CloudWatch, AWS WAF): Java/MySQL modernization, a from-scratch public API (OpenAPI), AI-powered document processing, an EDI-based carrier-reporting framework, and an offline encrypted enrollment app for restricted, no-internet environments.
• Built and operated a multi-year national serverless quiz platform for Università Bocconi (Cloud Run, Firestore, Flask, Angular) at near-zero idle cost.

Lead developer for an image-processing cloud-based system.

• Lead engineer for a contract-only image-processing cloud platform — customer-facing, onboarding teams, and managing up to 5 engineers; architected on domain-driven design, event sourcing, and CQRS.
• One of four core maintainers of Spine Event Engine, TeamDev's open-source Java DDD/CQRS/event-sourcing framework — commit, merge, release, and design-review ownership from 0.10.x to v2.alpha.
• Designed the product's gRPC/Protobuf contracts — .proto schemas, services, and server implementations in Java, Python, and Node.js — plus custom protoc plugins for multi-language codegen.
• Improved system architecture and infrastructure, leading to a 40% infrastructure cost decrease.

Built and maintained billing and ordering systems for A1 Telekom Austria.

• Built and maintained A1 Telekom Austria's ordering & billing system (Java 7→8) on Tomcat/EJB with Apache Camel and Spring, integrating the Amdocs process manager.